NASA Contractor Report 4189
High-Power AIGaAs Channeled
Substrate Planar Diode Lasers
for Spaceborne Communications
J. C. Connolly, B. Goldstein, G. N. Pultz,
S. E. Slavin, D. B. Carlin, and M. Ettenberg
CONTRACT NAS 1-17441
NOVEMBER 1988
11/36
Unclas
01_8960
- i
= ,
] 7[i
. =:: ....
2;_:_- q
2
i
:2;
• _2
' "gL-"
L=22
:.5!-_
=
==
=
_'_ _3
]]L
https://ntrs.nasa.gov/search.jsp?R=19890005055 2020-03-14T13:38:23+00:00Z
i.-. -
- -T-.-_-.__-.... -_..... _---
_ __--_- - __L
,.o __
• I----- T
_ _r_ -_
$
i
!
il
NASA Contractor Report 4189
High-Power AIGaAs Channeled
Substrate Planar Diode Lasers
for Spaceborne Communications
J. C. Connolly, B. Goldstein, G. N. Pultz,
S. E. Slavin, D. B. Carlin, and M. Ettenberg
David Sarnoff Research Center
Princeton, New Jersey
Prepared for
Langley Research Center
under Contract NAS1-17441
National Aeronautics
and Space Administration
Scientific and Technical
Information Division
1988
PREFACE
This report describes work performed from 28 June 1986 to 27 June 1987 at
RCA Laboratories in the Optoelectronics Research Laboratory, B. Hershenov,
Director, under Contract No. NAS1-17441. M. Ettenberg was the Project
Supervisor, and D. B. Carlin, J. C. Connolly, G. N. Pultz, and B. Goldstein were
Project Scientists. Other contributors to this research were J. K. Butler,
G. A. Evans, N. A. Dinkel, F. R. Elia, M. G. Harvey, D. B. Gilbert, T. R. Stewart,
J. J. Hughes, E. DePiano, D. P. Marinelli, D. T. Tarangioli, N. W. Carlson,
D. J. Channin, V. J. Masin, S. E. Slavin, F. Z. Hawrylo, S. L. Palfrey, and
A. R. Dholakia.
iii
aRECEDING PAGE BLANK NOT [_[LMED
Table of Contents
Section Page
PREFACE ...................................................................... iii
S_Y°°.°0°00,0°.,.°°°°,°°,°°°=°°o°°°°oi°.,°°.,iI°o°°,o°°°°°oi0o°i°°°°°° 1
Io 3
II. LASER PERFORMANCE AT 8650._ .................................
A. Computer Modeling ..................................................
B. Liquid Phase Epitaxial Growth ..................................
C. Laser Diode Processing ..............................................
D. Laser Diode Die and Wire Mounting ............................
E. Laser Diode Operating Characteristics ........................
F. Lifetesting and Reliability Assurance ..........................
G. Post-Life Failure Analysis ..........................................
4
4
13
25
27
29
33
36
III. LASER PERFORMANCE OF DFB-CSP STRUCTURE .........
A. Device Modeling ........................................................
B. Grating Fabrication and LPE/MOCVDGrowth Techniques ...................................................
C. Laser Diode Operating Characteristics ........................
39
39
41
48
IV. CONCLUSIONS ............................................................. 51
REFERENCES ................................................................ 52
APPENDICES
A. A Self-Consistent Analysis of Gain Saturation inChanneled-Substrate-Planar Double-HeterojunctionLasers .....................................................................
(Used by permission of Southern Methodist University, Dallas,TX 75275.)
55
v PRECEDING PAGE BLANK NOT FILMED
Table of Contents (cont'd.)
Section
B. Observations and Consequences of Non-UniformAluminum Concentrations in the Channel Regionsof A1GaAs Channeled-Substrate-Planar Lasers ............
(Used by permission of Solarex Corporation, Newtown, PA18940 and Southern Methodist University, Dallas, TX 75275.)
C. Effects of Ion Sputtering on the Surface Compositionof GaAs Laser Diode Facets ........................................
D. Intrusions in the Active Layer of Channeled-Substrate-Planar Lasers ............................................
E. A High-Power, Channeled-Substrate-Planar A1GaAsLaser .......................................................................
(Used by permission of Southern Methodist University, Dallas,TX 75275.)
F. An Efficient A1GaAs Channeled-Substrate-PlanarDistributed Feedback Laser ........................................
(Used by permission of Solarex Corporation, Newtown, PA
18940.)
Page
69
113
125
141
145
v±
List of Illustrations
Figure
.
.
,
.
,
.
.
.
.
10.
Complex lateral effective index for a 8650-/_ CSP laser as a
function of n-cladding layer thickness .................................
Complex lateral effective index profile for a 8650-/_ CSP
laser with an n-cladding layer thickness of 0.22 and 0.35
_Lm as a function of the V-channel width .............................
Normalized loss for a 8650-/_ CSP laser as a function of n-
cladding layer thickness ....................................................
Optical gain characteristics of the fundamental mode as a
function of the intracavity power. The drive current is for
a device of length L = 100 pm ..............................................
The intracavity power computed for a laser of length L =
250 pm. The back facet has R2 = 1, while the front facet
reflectivity is treated as a parameter. The total emission
power from the front facet is 50 mW ....................................
Emission power from the front facet of an A1GaAs laser vs
drive current ....................................................................
A photograph of an automated liquid phase epitaxial
growth reactor used for the growth of CSP lasers ..................
A photograph of the graphite growth boat used for the
fabrication of CSP lasers ....................................................
A schematic diagram and cross-sectional photograph of
the CSP laser structure .....................................................
(a) Geometry of a typical CSP type laser; x = 0 is the top of
the active layer and x = 1.8 /_m is the bottom of the
channel. (b) Auger analysis of a cleaved facet of a CSP
laser showing a higher aluminum composition near the
bottom of the channel (X = 1.4 pm, dashed line) than near
the top of the channel (x = 0.4 pm, solid line) ........................
Page
6
7
9
11
12
13
14
15
16
18
vii
List of Illustrations (contkL)
Figure
11.
12.
13.
14.
Index profiles (---) and corresponding electric field
distributions (--) for (a) a conventional CSP laser, (b) a
CSP-LOC laser; and (c) an ESL-CSP laser. The layer
compositions, thicknesses, and effective index for each
structure are listed in Tables 1-3. The dashed rectangles
in (a) and (b) show the field distributions on expanded
scales for x > 1.8 pm ..........................................................
(a) Calculated near-field FWHP as a function of the %
AlAs (or index of refraction at k = 0.83 _m) of a 0.4 pm
(---) and 0.9 pm (--) thick LOC layer (CSP-LOC
geometry) or of the 0.9 pm ( ...... ) thick n-cladding layer
(ESL-CSP geometry). The n-cladding layer for the CSP-
LOC layer has an AlAs mole fraction of 0.33, and the LOC
layer for the ESL-CSP has an AlAs mole fraction of 0.33.
(b) The calculated near-field FWHP as a function of AlAs
or index of refraction for the ESL-CSP structure on an
expanded scale. The common point to all three curves (at
an index value = 3.40657) corresponds to the conventional
CSP laser described in Table 1 ............................................
Composition measured by Auger analysis at four
positions along a cleaved facet of a CSP-type laser showing
a lower aluminum composition near the bottom of the V-
channel than near the top ..................................................
The substrate confinement factor, Fs, as a function of the
% AlAs (or index of refraction at k = 0.83 N m) of the n-
cladding layer at the bottom of the channel of an ESL-CSP
laser. The inset far-field intensity vs angle patterns show
a large variation in asymmetry as a function of % AlAs of
the n-cladding layer ..........................................................
Page
20
22
23
24
viii
List of Illustrations (cont'd.)
Figure
15.
16.
17.
18.
19.
20.
21.
22.
Schematic diagram of the combined electron beam
deposition and Auger analysis system use for facet
coating experimentation ....................................................
Removal rate of oxygen from the laser facet as a function
of ion sputtering time using 1000-eV argon ions ...................
Power output/current input (P-I) curves, spectrum, and
far-field radiation patterns at various output power levels
for a high-power CSP laser ................................................
Response of a CSP high-power laser to square current
pulses at 14% duty cycle. The fall and rise times are <0.5
ns .°,..°°..,°.°.°,°,.. °.°.,......°..,.° ,°,o, °.,o, °,......., °,.....,....,°......,o....
Aging behavior of high-power, 8600- to 8800-/_ CSP lasers
at 25°C as a function of operating time. The lasers were
maintained at a constant output power level of 50 mW
(50% duty cycle; 10 MHz) ....................................................
Aging behavior of high-power, 8600- to 8800-/_ CSP lasers
at 50°C as a function of operating time. The lasers were
maintained at a constant output power level of 50 mW
(50% duty cycle; 10 MHz) ....................................................
Aging behavior of high-power, 8600- to 8800-A CSP lasers
at 70°C as a function of operating time. The lasers were
maintained at a constant output power level of 50 mW
(50% duty cycle; 10 MHz) ....................................................
(a) Pre-lifetest, near-field pattern and light-intensity scan
for a high-power CSP laser. (b) Post-lifetest, near-field
pattern and light-intensity scan for the same CSP laser ........
Page
26
27
3O
32
34
35
36
38
ix
List of Illustrations (cont'd.)
Figure
23.
4.
25.
26.
27.
28.
29.
(a) Schematic diagram of CSP-DFB laser. (b) Stained
cross-sectional cleave of CSP-DFB structure. (c) Stained
cross-sectional cleave lapped at a 1 ° angle in the vertical
direction. Note especially the beginning of meltback
between the n-cladding and n-buffer layers ..........................
Photograph of metalorganic chemical vapor deposition
(MOCVD) system used for the epitaxial growth of the DFB-
CSP laser .........................................................................
A grating with a 2400-/_ period formed in an A1GaAs
layer by chemical etching techniques ..................................
A grating with a 2400-/_ period formed in an A1GaAs by
ion-beam milling techniques ..............................................
(a) An SEM cross-sectional photograph of an ion-beam-
milled grating in a GaAs substrate. (b) An SEM cross-
sectional photograph of an ion-beam-milled grating in
14% AlAs with a 40% AlAs layer grown on top by
MOCVD. (c) An SEM cross-sectional photograph of an
ion-beam-milled grating in 23% AlAs with GaAs grown
on top by MOCVD ..............................................................
Emission wavelength shift as a function of heatsink
temperature for a CSP-DFB laser operating pulsed at an
output power of 10 mW ......................................................
(a) Power-current curves for a CSP-DFB laser. (b) Far-
field radiation patterns for a CSP-DFB laser .........................
Page
42
43
45
45
47
48
49
SI._IMARY
The fabrication procedures and theoretical understanding of high-power,
8600- to 8800-/_ channeled substrate planar (CSP) lasers have been expanded,
particularly in areas that focus on increasing power capability and reliability.
These improvements have been realized without sacrificing the superior
properties of the CSP laser, such as non-astigmatic wavefronts, modulation
performance, and beam quality.
The single-spatial-mode, output power level for a discrete device hasreached 70 mW cw. This value was chosen because of "kink" in the P-I curve and
broadening of lateral far-field at values above 70 mW. The overall power capability
for the laser has been extended to 190 mW cw, a 50% improvement over the results
reported previously. The typical lateral and perpendicular far-field radiation
patterns at the beam full width half power (FWHP) point for these devices are 7°
and 27°, respectively. Although no means are provided to stabilize the
longitudinal mode of the laser, a few selected devices displayed stable single mode
cw operation at power levels up to 90 mW.
The results of computer modeling studies identified the importance of
optical absorption or loss in the CSP structure and we have been able to correlate
these findings with experimental results. In addition, we have found that the
thickness of the n-cladding layer in the structure does not significantly alter the
lateral effective index profile. In other related studies, we were able to identify
and correlate the effect of compositional changes on the structure.
The reliability of the CSP lasers has also benefited from our new findings.We have been able to increase the lifetesting power level from 30 to 50 mW and
have changed our test mode from a constant-current to constant-power format.
This format change has permitted us to increase the stress level placed upon the
devices and to monitor the performance of the devices in a mode similar to their
operation in space. The results of CSP lasers placed on lifetest at 50 mW (50%
duty cycle; 10 MHz) output power and at operating temperatures of 25°C, 50°C,
and 70°C have shown room temperature lifetimes of approximately 3,000 h, a
significant improvement over our previous work.
A new type of CSP structure has also been demonstrated, the distributed-feedback (DFB) CSP laser. This structure contains a grating that stabilizes the
longitudinal mode of the CSP laser. Distributed-feedback operation has been
obtained at room temperature over an 8°C temperature range. This structure has
2
not only provided longitudinal mode stability but has displayed superior
wavelength-temperature dependence over the conventional CSP structure (0.7 vs 3
/_J°C) without degrading its excellent performance properties.
HIGHLIGHTS
High-power, single-spatlal mode 8600- to 8800-/_ CSP lasers
• 190-mW, cw output power capability
• Single spatial mode operation up to 70 mW cw
• Room temperature lifetimes of 3,000 h at 50 mW
• Improved computer modeling capabilities
• Modulation rate to 2 GHz
High-power, distrlbuted-feedback CSP lasers
• Demonstration of first DFB-CSP laser structure
• Single longitudinal mode operation up to 40 mW
• 0.7/_/°C wavelength dependence upon temperature
I. INTRODUCTION
The work described in the previous Annual Report centered on the
development of both individual and arrays of high-power, single-mode diode
lasers for potential use in areas such as space communications, optical data and
storage, and local area optical communication networks. In the work on
individual laser sources, we reported record power in a fundamental mode from
channeled substrate planar (CSP) lasers having improved efficiency and reduced
threshold current density without sacrificing the excellent beam qualities of the
device. Similar results were reported for the individually addressed CSP laser
diode arrays.
In this annual report, the research and development work was continued to
improve further and refine the high-power, CSP laser structure for wavelength
emission at 8600 to 8800 _. Extensive computer modeling studies were conducted
to identify the ultimate output power, performance, and reliability limitations of
the device as well as the parameters that control them. The results of these
studies were used to fabricate CSP diode lasers that have displayed higher output
powers, lower threshold currents, higher efficiencies, and improved reliability.
A new laser structure was also developed that incorporates a grating to
stabilize the longitudinal mode of the CSP diode laser during modulation.
Another added benefit of the grating is the improved wavelength temperature
dependence over conventional devices. The improvements described above have
been shaped by the requirements of the NASA Advanced Communications Test
Satellite (ACTS) program.
H. DIODE LASER PERFORMANCE AT 8600- TO 8800-A
The goal of this program was the development of A1GaAs diode lasers with
an emission wavelength at 8600 to 8800 /_ comparable in performance with
previous diode lasers fabricated at 8350/_. In addition, the focus of the program
was to develop a new diode laser structure that operates in a stable longitudinal
mode. The standard diode laser geometry that provides operation in a stable,
single-spatial mode offers no method of longitudinal mode stabilization. The
eventual use of these diode lasers would be as light sources for use in intersatellite
communications systems and, specifically, the NASA Advanced Communica-
tions Technology Satellite (ACTS) System [1].
In this section, computer modeling, liquid phase epitaxial (LPE) growth,
characterization, and reliability of the 8600 to 8800/_, high-power CSP diode lasers
are discussed. A later section describes the method used to incorporate a grating
in this structure to stabilize the longitudinal mode without degrading its desirable
performance characteristics.
A. COMPUTER MODELING
The computer modeling program used at the David Sarnoff Research
Center for this work has been developed and refined over years as our
understanding of the operation of the CSP structure has improved. The method of
modeling that we used starts with the basic laser structure and the various
components associated with laser operation, such as the current distribution,
carrier diffusion gain profile, and heating effects, which are sequentially
incorporated into our model. As these various components are added, iterations
are performed until self-consistency is obtained within the error limits imposed
upon the device parameters.
A close examination of both the structure and the performance
characteristics of high-power CSP lasers grown in the three LPE reactors used for
this work was initiated. The goal of this study was to identify those parameters
that yielded the most significant consequence to the high-power performance and
operational lifetime of the CSP diode laser.
The first experimental parameter identified as significant to the
performance of CSP lasers operating at an emission wavelength >8600/_ was the
thickness of the n-cladding layer. Although no correlation could be found between
the thickness of the n-cladding layer and high-power operation, there was an
indication that CSP structures containing thicker than normal n-cladding layers
(0.4 to 0.5 _m vs 0.2 to 0.3 _m) displayed longer operational lifetimes. In addition,
these results were also supported by similar findings on two other A1GaAs LPE
reactors used for the growth of phase-locked arrays and 0.83-pm CSP lasers. Thisresult may be significant since the increased operational lifetimes may be related
to the design of the CSP structure itself and not with the quality of the LPE
prepared materials, processing, and/or the growth system.
Based upon these results, a computer modeling study was performed
varying the thickness of the n-cladding layer to determine the effect on the lateral
index, gain, and optical field profiles in the CSP structure. In addition, the totalmodal losses were examined and found to be insensitive to the n-cladding
thickness. These losses would directly effect the threshold current value of thedevice. An increase in the modal loss would lead to a corresponding increase in
threshold current due to the higher gain necessary to support lasing. The
experimental results support these findings since no correlation between the n-
cladding thickness and threshold current was found.
In our modeling, we use the effective complex index technique to calculate
any change in the mode properties in the laser structure [3] that would alter the
CSP lasing properties. The complex effective index (n*) consists of a real and an
imaginary part and is defined as:
n* = Mk 0 = B/k 0 + j ¢x/ko
where [_ is the propagation constant, k 0 is free space wavenumber (2rdk), a is the
absorption coefficient, and _. = _ + ja is the longitudinal propagation constant of
the mode. The complex effective index is calculated both in the channel and the
winged regions of the structure, and the difference between these values is called
the lateral effective-index step. The lateral effective-index step (An) in the CSP
structure is defined as:
An = _/k 0 (channeled region) - _/k 0 (winged region)
Figure 1 shows a plot of the complex lateral effective index as the thickness
of the n-cladding layer is varied between 0.2 and 1.5 _m is shown. The structure
was modeled using an n- and p-cladding layer composition of Alo.27Gao.73As and
3.440
XWaz
w
I--UImliiimiii
3.435
3.430
3.425
n- and p-clld : 27% AI
active layer - 1% AI; 600 A
3.420 , , , , , , , , • , ,
0.0 0.2 0.4 0.5 0.0 1.0 1.2 1.4 1.6 ! .8 2.0
Figure 1. Complex lateral effective index for a 8650-,/k CSP laser as a function of
n-cladding layer thickness.
an active layer composition of Alo.olGao.99As. The active- and cladding-layer
thicknesses are 0.06 and 1.5 _m, respectively, and the channel width is 4 _m. For
an n-cladding layer thickness greater than 1 _m, the effective index remains
unchanged at 3.436 (i.e., channel region) and decreases to 3.424 for a thickness of
0.25 _m. Thus, the lateral effective index step for a CSP laser with a n-cladding
layer thickness of 0.25 _m is 1.2 x 10-2. The steep slope of the curve for n-cladding
thicknesses between 0.25 and 0.6 pm demonstrates the uniformity and thickness
control necessary for the growth of this layer. Small changes in the thickness of
this layer can significantly alter the performance of the CSP laser..4360
XUJaz
IJJ>I--(JLIJU_U_UJ
Figure 2.
3.4355
3.4345
\n-CLAD = 0.22 um
n- Ind p-clnd = 27% AI
active layer - I%AI; 600 A
3.4330.
2 :3 4 5 6
CHANNEL WIDTH (urn)
Complex lateral effective index profile for a 8650/_ CSP laser with an n-cladding layer thicknesses of 0.22 and 0.35 _m as a function of the V-channel width.
In Fig. 2, a plot of the complex, lateral effective index for CSP laser
structures with n-cladding layer thicknesses of 0.22 and 0.35 pm is shown. The
other parameters for the CSP laser remain unchanged. As the channel width is
varied from 2 to 6 _m, the complex lateral effective index profiles for the two
different n-cladding layer thicknesses varies at most 5.5 x 10-4for the range of
channels (i.e., 3.0 - 4.5 _m) used in fabrication. This small index difference
would not significantly alter the operation of the CSP structure. Generally, a
change of 2 x 10-3 or greater is required to change the modal properties of thestructure.
Although the total modal losses in the CSP structure were found to be
insensitive to the n-cladding thickness, the normalized loss or the loss parallel tothe direction of the active layer varied significantly. In Fig. 3, the normalized loss
for the CSP structure is plotted against the n-cladding thickness. The device
parameters are the same as used previously. There is a significant change in the
amount of loss (8 x 10-3) as the n-cladding layer thickness is reduced from 1.5 to
0.2 _m. Increasing the thickness to 0.4 _m reduces this loss to 3 x 10-3. Loss in
the CSP structure is critical for its operation in a single-spatial mode. A small
change in loss represents a large change in the complex lateral effective index
profile. If the loss value was zero in the structure while the real part of complex
lateral index profile remained unchanged, the structure would behave as a ridge-guide-type laser.
The loss mechanism in the CSP structure is characterized by absorption of
the optical mode by the highly absorbent GaAs substrate (10,000 cm-1). The optical
absorption in the substrate causes local heating of the substrate on either side ofthe V-channel. This heat must be removed from the substrate for the CSP laser to
operate at high powers with reliable lifetimes. In our normal mountingconfiguration, the CSP laser is mounted p-side down to the heatsink. Thus, the
heat in the substrate must be removed by re-radiation of the absorbed heat across
the active region, through the p-cladding and cap layers, to the copper heatsink.
Absorption cannot be totally eliminated in the CSP laser since absorption of the
optical mode by the substrate is crucial to operation in a single-spatial mode at
high output powers.
9
O)00O
C_wN
n,oz
0.008 -
0.006
0.004
0.002
n- and p-clad = 27_IAI
active layer - I_I AI; 60OA
Figure 3. Normalized loss for a 8650-/_ CSP laser as a function of n-cladding
layer thickness.
The theory described above that associates long operational lifetimes with
reduced heating along the shoulders of the V-channel fits the experimental
evidence found from our analysis of lifetest data. The computer model used for
this study clearly demonstrates that thicker n-cladding layers reduce the localized
loss or absorption of the optical mode by the GaAs substrate. This reduction
lO
translates to reduced heating in and around the active region of the device. The
reduced heating permits the device to operate at a lower junction temperature
while still maintaining the same output power level. Thus, longer operational
lifetimes should be obtained. The absorption of the optical mode by the substrate
may prove to be a very important factor not only for increased reliability, but for
operation at higher output power levels.
In another modeling study, we investigated the effects of non-uniform
photon densities along the propagation direction for high-power CSP lasers with
low facet reflectivities. CSP lasers with high reflectivity facets (>30%) have almost
uniform photon density along their longitudinal axis, allowing the assumption of
uniform gain to be applied in their analysis. However, the highest powers
obtained from CSP lasers are obtained by using a high-reflectivity rear facet (80%-
90%) and a low-reflectivity output facet (0.5%-10%), resulting in large variations of
the photon density along the lasing axis. By extending our previous work on
uniform longitudinal gain analysis based on the "self-consistent model" [3], we
have been able to develop simple phenomenological equations for the modal gain
coefficient, the threshold current density and, in the limit of high facet
reflectivities, the radiated power.
In self-consistent models the carrier density in the active layer is derived
from a solution of the diffusion equation having both source and sink terms. The
source for the injected carriers is the drive current, whereas the sink is the
stimulated recombination. The current flow into the active layer varies laterally,
and the lateral carrier diffusion within the active region affects the optical gain
profile, modifying the shape of the optical field distribution in the CSP laser diode.
The spatial dependence of the recombination term is computed from the product
of the lateral gain profile and the photon density in the active layer as a function of
position along the longitudinal direction.
In any laser structure the intracavity power P(z) is the sum of the forward
and backward waves and is defined as:
P(z) = P+(z) + P-(z)
For lasers having high facet reflectivities (>30%), the intracavity power is almost
constant along the z, or lasing, direction. However, when reflectivities are lower,
as in the case for high-power CSP lasers, the intracavity power exhibits relatively
large changes along the lasing direction. In Fig. 4, the points show the modal
11
gain coefficient, computed from the self-consistent solutions of the carrier
diffusion and Maxwell equations, as a function of the intracavity power for
different values of drive current. The CSP device used for this work contained a
0.06-_m-thick GaAs active layer having an index of 3.6, a 0.4-_m-thick A1GaAs n-
cladding layer having an index of 3.4, and a 1.5-_m-thick A1GaAs p-cladding
layer having an index of 3.4. Both the V-channel and contact stripe widths were
6 _tm.
120q_,
100_
:00s0
r.D
G
=
"_°_oz
_'---.__ _,_. I
-2o I I I I20 40 60 80 100
INTRACAVITYPOWER,P (roW)
Figure 4. Optical gain characteristics of the fundamental mode as a function ofthe intracavity power. The drive current is for a device of lengthL = 100 pro.
Although the numerical data can be used to calculate the longitudinal
variation of the gain in a laser with known facet reflectivities, it is useful to fit
numerically calculated modal gain coefficient points to an analytical expression
written as:
o0G(P) = (1 + P/Ps) d " az
where c, d, GO, Ps, and al are unknown constants. An optimization procedure for
a least-squares fit to the computed values of the self-consistent modal gives c =
12
0.708, d = 0.687, GO = 51.4 cm -1, Ps = 41.2 mW, and al = 49.6 cm -1. The value of Io is
arbitrary, but we used Io=10 m_A for the GO value above. The resulting gain curves
for these parameters are illustrated in Fig. 4.
When one or both of the facet reflectivities are small, the intracavity power
varies considerably along the lasing axis. Thus, the intracavity power must be
computed from the integral equation
P(z) = Po exp G(P) dz' +_22exp z- _G(P) dz'o
where P0 is an eigenvalue, facet 2 lies at z = 0, and facet 1 lies at z = L. The
boundary condition on G(P) requires that its integral over the length of the cavity
be 1/21n(1/R1R2).
In Fig. 5, the results for a CSP diode laser operating at an output power
level of 50 mW and having a cavity length of 250 _m are shown. The rear facet of
the laser has a reflectivity of 100%, and the output facet reflectivity is treated as a
variable. When the output facet reflectivity is 30%, the power along the lasing axis
is almost constant. However, if the reflectivity of the output facet is reduced to 5%
by applying a 1/4 _. coating of Al203, the intracavity power varies from about 24
mW at the rear facet to about 55 mW at the emitting facet.100
R_=I.090
80E
_- 70f.aul
60
"_ S0a,e
== 40
30
R1=0.3
20 . I I I I 1 l I l I25 50 75 100 125 150 175 200 225 250
Figure 5.
AXIALPOSITION,Z(pro)The intracavity power computed for a laser of length L = 250 pm. The
back facet has R2 = 1, while the front facet reflectivity is treated as aparameter. The total emission power from the front facet is 50 mW.
13
A more useful way of looking at the effects of mirror reflectivity upon the
performance of the CSP laser is to examine the change in the output power vs
drive current (P-I) plots for the same laser diode having various output facet
reflectivities. Using the same device parameters as before, the P-I curves were
calculated for the various facet reflectivities and are displayed in Fig. 6. From the
figure, it can be seen that the threshold current value for the laser almost doubles
as the facet reflectivity is reduced from 30% to 5%. The slope efficiency of the CSP
laser also increases as the facet reflectivity is decreased. A more detailed
description of these modeling results can be found in Appendix A of this report.
50
40
E
30
zO
_, 20ms
R2:I.O
25 50 75
DRIVE CURRENT (mA)
10
I I100 125 150
Figure 6. Emission power from the front facet of an A1GaAs laser vs drivecurrent.
B. LIQUID PHASE EPITAXIAL GROWTH
The results of our computer modeling clearly points out the importance of
layer thickness, compositional, and layer uniformity across the growth substrate.
Thus, we introduced the use of a new, fully automated LPE growth system. This
LPE growth system was designed and constructed in-house since commercially
available systems did not contain the necessary modifications and the degree of
automation required for our type of growth process. Some of the unique features
associated with our growth system are: (1) all gas lines and system components
are fabricated from #316L stainless steel and are tungsten in gas (TIG) welded
wherever possible and metal gasket sealed when not; (2) a vacuum pumping
14
system has been added to the gas-handling system to permit a pump/purge
procedure to be performed on the growth chamber after loading of the growthboat; (3) automatic in situ temperature profiling of the growth boat during the
growth process; (4) automatic positioning of both the top and bottom sliders in the
growth boat; (5) automatic positioning of the three zone growth furnace; and (6)
continuous monitoring and control of the growth temperature via a thermocouple
mounted inside the growth boat. A photograph of this new reactor is shown in
Fig. 7. In addition, all aspects of the growth process (i.e., pressure, temperature,
gas flow rate, etc.) are continuously monitored and recorded. This system is the
most automated and sophisticated LPE growth reactor that we are aware of in theworld.
Figure 7. A photograph ofan automated liquidphase epitaxialgrowth reactorused
forthe growth ofCSP lasers.
In conjunction with the design of the new fully automated LPE reactor, we
have redesigned the LPE growth boat. It is still fabricated from ultra-high-purity
ORIGINAL PAGE IS
OF POOR QUALITY
15
graphite, but it has been modified to permit easy disassembly and cleaning along
with improved wiping action for more complete melt removal to reduce gallium
carry-over. In addition, the boat has been redesigned to accept larger substrates.
A photograph of the new LPE growth boat is displayed in Fig.8. The new
substrate size is 1.0 in. x 1.25 in. This increased size provides approximately 3.3
times more useable wafer area over our previous substrates. The growth melt for
this boat is 10 gm as compared with 3 gm for the older design.
Figure 8. A photograph ofthegraphitegrowth boat used forthe fab_cation of CSPlasers.
One of the most important considerations in fabricating the CSP laser
structure is the growth of a planar active layer while still maintaining a thin and
reproducible n-cladding layer. A schematic diagram and a cross-sectional
photograph of the CSP structure is shown in Fig. 9. If the active layer is
nonplanar, due to incomplete filling of the V-channel in the substrate while
trying to obtain a thin n-cladding layer, the lateral index profile of the CSP
structure is altered leading to devices exhibiting small lasing spots and single-
spatial mode operation at low power levels. On the other hand, if a planar active
layer is obtained but the cladding layer is too thick, the devices will behave as
gain-guided or oxide-defined contact stripe devices. The growth of a planar active
layer in conjunction with a thin (0.2 to 0.4 _m) n-cladding layer requires that the
growth melt be in a supersaturated condition prior to epitaxial growth. A
16
supersaturated condition is obtained by cooling a melt from an equilibrium
condition to a temperature above the spontaneous or self-nucleation temperature.
When the melt is in this state, it is normally referred to as supercooled. The exact
temperature of all three parameters depends upon the composition of the melt
itself. To ensure planar active-layer growth over the channel region while
maintaining the appropriate thickness n-cladding layer, it is necessary to
maximize the total amount of supercooling associated with the melt used for layer
growth. This is accomplished in our automated growth system by using the
single-phase growth method [4] for the n-cladding layer and the two-phase
method [4] for the growth of the other layers in the CSP structure. The single-
phase growth method permits us to accurately control the degree of supercooling
present in the gallium melt prior to the introduction of the growth substrate. In
addition to the quick filling of the V-channel to planarize the layer, careful control
of the amount of supercooling also permits a high degree of control on the layer
thickness, not only across the wafer, but from one LPE growth run to another.
The single-phase growth technique may also be used for the growth of the other
layers in the structure but is not necessary since the n-cladding layer grown
directly on the non-planar substrate planarizes the surface for growth of the
subsequent layers. The use of the single-phase growth technique for all the layers
in the structure would unnecessarily complicate the growth process.
F--ZINC DIFFUSION
g(.://///(_/(((///////A _ J P-CONFINEMENT._, -ACTIVE_ N-GONFINEWENT_'
----SUBSTRATE
Figure 9.
H5p.rn
A schematic diagram and cross-sectional photograph of the CSP laserstructure.
A detailed analysis of compositional variations within the layers of the CSP
structure has been performed, and compositional changes in the n-cladding layer
within the channel region have been observed. We have found that significant
non-uniformities in the direction perpendicular to the junction can exist in the
AUGa ratio. As a consequence, a large optical cavity (LOC) or enhanced substrate
17
loss (ESL) version of the CSP geometry may result, both of which may havesignificantly different characteristics from those of a conventional CSP laser. The
CSP-LOC laser generally has a wider perpendicular full-width-half-power
(FWHP) near-field distribution and similar or larger perpendicular far-field beamdivergence compared with a conventional CSP laser. The ESL-CSP laser has an
asymmetric perpendicular far-field pattern and can have either a larger or
smaller FWHP perpendicular far-field pattern compared with a conventional CSPlaser.
The principal experimental technique used for the surface compositional
analysis on the cleaved facets of the lasers was Auger electron spectroscopy (AES)
using a primary electron beam with a resolution of about 1000/_. Figure 10(b)shows the Auger spectra that indicate the surface composition on a cleaved facet
of a CSP laser at the points x = 0.3 _m and x = 1.2 pm shown in Fig. 10(a). The
magnitudes of the Ga, As, and A1 lines shown on the spectra reflect the
concentration of these constituents at the two points and indicate that the A1
concentration at the bottom of the channel is about twice that just below the active
region. Note that the change in the magnitude of the A1 line is tracked by a
corresponding change in the magnitude of the Ga line, while the As line has
remained essentially unchanged. Examining a random sampling of CSP lasers
establishes that the magnitude of the concentration variations indicated in
Fig. 10(b) ranges from zero to a factor of about two.
18
REGION B I REGION A I REGION B
CAP 0% ALAs
p-CLAD
ACTIVE JL7% AI.As m--_-- ..........
O OSp.m J _
2-
SUBSTRATE
33% ALAs
, L , * I , = , = I
t 2
(Fm)
0% AlAs 7
~ 75-1OO/_m_
(a)
O.5Fm
I Op.m
Y(F rn)
O__.3/_m
I_ t At 20g tlI
i As 53%. k= f
24 0
22 _
900 t000 I100 1200 1300 1400 1500
eV (AUGER ELECTRON ENERGY) 6537
Figure 10. (a) Geometry of a typical CSP type laser; x = 0 is the top of the active
layer and x = 1.8 lim is the bottom of the channel. (b) Auger analysis ofa cleaved facet of a CSP laser showing a higher aluminum compositionnear the bottom of the channel (X = 1.4 lim, dashed line) than near the
top of the channel (x = 0.4 _m, solid line).
19
A non-uniform A1/Ga ratio within the channel will affect the dielectric
profile perpendicular to the junction. Figure 11 contains a series of possible index
profiles together with their electric field distributions for (a) a uniform A1/Ga ratio
in the channel, (b) a higher A1/Ga ratio at the channel bottom (CSP-LOC), and (c)
a lower A1/Ga ratio at the channel bottom (ESL-CSP). Graded-index profiles are
also possible; however, their characteristics are qualitatively similar to the abrupt
step profiles. The A1 concentrations and corresponding index values for the three
structures shown in Fig. 11 are tabulated in Tables 1, 2, and 3.
Table 1. CSP Structure.
layer thickness (prn) %AlAs index (k=0.83pm) Flayer
1 p-clad >1.0 33 3.40657 .389
2 active 0.08 7 3.62805 .222
3 n-clad 1.8 33 3.40657 .389
4 substrate - 75 0 3.64 :3.4 x 10 -6
Re[n'el f} lm{n*eff }
3.43401 6.6 x 10 -5
Table 2. CSP-LOC Structttre.
layer thickness (p.m) %AlAs index (k=0.831am) Flayer
1 p-clad > 1.0 33 3.40657 .19043
2 active 0.08 7 3.62805 .19205
3 LOC 0.4 22 3.48276 .59205
4 n-clad 1.8 40 3.364 .02546
5 substrate - 75 0 3.64 2.9 x 10-9
Table 3. ESL-CSP Structure.
Re{n'el f}
3.46932
lm{n*eff}
6.6 x 10 -5
layer thickness (l-tin) %AlAs index 0,.=0.831am) Flayer
1 p-clad > 1.0 33 3.40657 .331
2 active 0.08 7 3.62805 .190
3 LOC 0.9 33 3.40657 .346
4 n-clad 1.8 26.3 3.45245 .125
5 substrate - 75 0 3.64 .008
Re {n'eft}
3.43405
Im{ n*eff}
5.11 X 10-4
20
2.0
T,srE 10'£.)
E:
_ .0.5i,i1W
0.0
I 2
I Tw 1
° °I_
' I ' I ' I ' I ' I '
-0.5 3.3-2 3
X (/_m)
Figure 11(a)
4 ' I ' ] ' i ' ] ' I '
JuJ 2
uio_ o I
J -2
g_I I I I I , I , I , I
_ 4t
1.8 2.0 2.2 2.4 2.6 2.8 3.0I
X (pro) I
2.0 j,i , ;; ,I,,, ,, , i, , ,,,, ,, ,, 1_,]13.?
,.s = i s.6n-CLAD _
,T I.o - s.5n,.
0.0 ..... j H
-0.5 I-II' II '+ '. I' ' '' I' 'I ' I j '''l'l i' I 'l ' 32-I.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5
X (/_m)
_gm'e 11(b)
21
2.0
1.5
w
u_ 1.0o
a-
ow 0.5
w
n. 00
'' ''I''''I''''I''' 'I''''I''''-0 ® ® ® ® -
eSL_ ESL
LOCI n CLAD
= B� .... _
-2 -1 0 I 2
X (H.m)
3,7
z5.6 o
I-
3.5w11:
5.4 oxw
3.3 zH
3.2
Figure 11(c)
Figure 11. Index profiles (---) and corresponding electric field distributions (m)
for (a) a conventional CSP laser, (b) a CSP-LOC laser; and (c) an ESL-
CSP laser. The layer compositions, thicknesses, and effective index
for each structure are listed in Tables 1-3. The dashed rectangles in
(a) and (b) show the field distributions on expanded scales for x > 1.8
_lm.
The data in Fig. 10 indicates that the aluminum concentration is highest at
the bottom of the channels. For this case, the computed effective perpendicular
index profile is no longer that of a simple double-heterostructure (DH); it is that of
a large optical cavity (LOC). Such a dielectric profile was introduced intentionally
to lower the optical power density by producing a wider perpendicular near-field
distribution than that of a conventional DH configuration [5]. In Fig. 12, the
FWHP of the near-field intensity is plotted as a function of AlAs composition (or
index at k = 0.83 pm) of the LOC layer assuming the AlAs mole fraction of the p-
and n-cladding layer is 0.33 and that of the active layer is 0.07. The CSP-LOC
FWHP near-field intensity of the fundamental mode reaches a maximum of about
0.4 pm (for a 0.4-_m-thick LOC layer) and 0.72 pm (for a 0.9-pm-thick LOC layer)
at AlAs compositions of the LOC layer of about 18%. These curves are designed to
indicate trends in behavior of the mode and do not consider, for example, higher
order modes, which are possible in thick LOC layers.
22
F-0.8
zLIJ
0.7zH
a 0.6Jhi
Li_ A
rr E 0.5w
z 0.4u.I"1-t-
,, 0.30
CL-r 0.2
u..
% ALAs
30 25 20 15 I0I I I I I
0.9Fm LOC --
....... ............i.40 3.45 3.50 3.55 3.60
INDEX OF LOC LAYER (CSP-LOC) ORESL n-CLAD LAYER (ESL-CSP)
0I-
1
3.65
>-F--
zL_
ZI-4
C_.J
u. Er,-:L
u.Jz
w-l-
b..0
0.25625
0.25600
0.25575
0.25550
0.25525 -
0.25500 , I3.40
3OI
•j
% ALAs
25 20 15 10I 1 I I
_[]--r_s L -CSP
• "°1 /
°• ••..,•o,=°•
,, I, ,, ,I ,,,,I,,,,I,,,,5.45 3.50 3.55 3.60 5.65
INDEX OF ESL n-CLAD LAYER (ESL-CSP) 6540
Figure 12. (a) Calculated near-field FWHP as a function of the % AlAs (or index
of refraction at % = 0.83 pm) of a 0.4 pm (---) and 0.9 pm (_) thick
LOC layer (CSP-LOC geometry) or of the 0.9 pm ( ...... ) thick n-
cladding layer (ESL-CSP geometry). The n-cladding layer for the
CSP-LOC layer has an AlAs mole fraction of 0.33, and the LOC layerfor the ESL-CSP has an AlAs mole fraction of 0.33. (b) The calculated
near-field FWHP as a function of AlAs or index of refraction for the
ESL-CSP structure on an expanded scale. The common point to all
three curves (at an index value = 3.40657) corresponds to the
conventional CSP laser described in Table 1.
23
CAP
p- CLADACTIVE ---_,-,
n -CLADQ4
o1
SUBSTRATE
FACETPOSITION
1
2
3
4
COMPOSITION% AL % Ga % As
0 5O 50
14 7,5 53
20 29 52
18 _,2 52
Figure 13. Composition measured by Auger analysis at four positions along acleaved facet of a CSP-type laser showing a lower aluminumcomposition near the bottom of the V-channel than near the top.
Some CSP structures examined also displayed lower A1 concentrations at
the bottom of the channel, which is opposite to our previous findings. Figure 1 3
shows a schematic drawing of the channel region of a CSP laser together with the
A1 concentrations, as given by AES data. The effective perpendicular index
profile, displayed in Fig. 14, can be thought of as pulling some of the optical mode
power into the substrate, thereby increasing the mode loss. This mechanism,
which we call Enhanced Substrate Loss (ESL), can also be explained by realizing
that all the modes supported by the index profiles are complex modes (i.e., the
longitudinal and transverse wave vectors have a real and an imaginary
component) because the field solutions to the electromagnetic wave equation [6]
are sinusoidal in the substrate. Usually, the n-cladding layer separating the
active region from the substrate is thick enough (>1 _m) that field penetration of
the laser mode into the substrate is negligible. However, if the mole fraction of
AlAs is lower at the bottom of the channel than at the top of the channel, the
higher index portion in the channel acts as an anti-reflection coating between the
high-aluminum, low-index portion and the no-aluminum, very-high-index
substrate, thereby coupling or redistributing a larger fraction of the mode power
into the substrate.
24
{D
0.008
0.006
0,004
0.002
-50 I
0.0003.40
- 50 0 50
3.45
% ALAs A
-5o.... o .... _ i/
4o" "'i'0 / A
I I I 1 I I I illl_ll_{I II I3.50 3.55 360 3.65
INDEX OF REFRACTION OF THE ESL n-CLAD LAYER
Figure 14. The substrate confinement factor, r s, as a function of the % AlAs (orindex of refraction at t = 0.83 ttm) of the n-cladding layer at the bottomof the channel of an ESL-CSP laser. The inset far-field intensity vsangle patterns show a large variation in asymmetry as a function of %AlAs of the n-cladding layer.
The exact cause of a non-uniform AUGa ratio occurring within the channel
region of a CSP laser during the growth of the first cladding layer is not fully
understood. However, we do know that (1) layer growth must be faster at the
bottom of the channel than at the top to obtain a planar surface, (2) both lateral
and perpendicular growth components must be present within the channel, (3)
the sidewall and bottom of the channel present different crystallographic planes
for nucleation that are different from those above the channel, where the growth
is planar, and 4) varying degrees of meltback may be present at the sidewalls and
bottom of the channel. Any of these conditions can readily affect the composition
of the ternary compound that initially nucleates and freezes out from the A1GaAs
melt, and can alter this composition as the growth proceeds and overall growth
conditions change. Furthermore, if the A1/Ga ratio changes during the initial
growth, then changes in the local melt composition may occur, which might
further change the A1/Ga ratio later in the growth process. Data from the AlAs-
GaAs phase diagram shows that very small changes in the A1 content in the melt
produces very significant changes to the A1 content in the grown material. The
25
basis for a non-uniform A1/Ga ratio within the channel of a CSP laser is also
consistent with the variability of this effect from channel to channel since it would
be the local conditions around each channel that would determine the magnitude
of the effect. A more complete description of these abnormalities can be found in
Appendix B of this report.
C. LASER DIODE PROCESSING
In the wafer processing area, all aspects associated with the different
processes (i.e., diffusion, metallization, photolithography, etc.) are closely
monitored to maintain a high throughput yield. In addition, the processes are
closely scrutinized periodically and changed as the need arises for improvements
in both yield and process quality. One area that we examined was the effects of
ion sputtering or milling of the emitting facets of the CSP lasers. The cleaved
mirror facets of the CSP lasers contain native oxides, owing to the reactive nature
of the aluminum alloys used in the growth of the structure and absorbed
impurities (i.e., water, carbon, etc.) on the surface caused by various fabrication
procedures. A pre-deposition cleaning using argon ion-beam milling should
increase adhesion of the facet coatings (a problem that has plagued this process
for years) and improve device reliability. The principal problem associated with
the use of this milling or sputtering process is the damage that may result at the
mirror facet of the laser. The exact point at which this cleaning procedure is a
benefit and at which damage to the laser facets occurs was unknown. As a
result, we implemented a study to determine all the variables associated with this
process.
In this study, we used AES to quantify changes in the surface composition
of the laser facets generated by the argon ion bombardment. Deviations from
stoichiometry and surface-oxide removal rates were examined for argon ion
current densities between 0.02 and 0.04 mA/cm 2 and ion energies in the range of
200 and 1000 eV.
The experiment was performed in a conventional cryo-pumped electron
beam deposition system having a base pressure of 10 -7 Torr. The ion source used
was a broad beam type, and the physical constraints within the deposition
chamber resulted in placement of the ion gun at an angle of 5 ° off the normal of
the sample surface. The argon flow rate to the ion source was adjusted via a
mass flow controller to yield an operating pressure of 10 -4 Torr during the ion
26
milling process. A biased plasma probe with a grounded shield could be
interchanged with the sample, thus enabling in situ measurements of the ion
current density prior to actual sample exposure to the beam.
An AES system is attached to the electron beam system and kept isolated
during the deposition process by a gate valve. A magnetically coupled
feedthrough is used to transfer the samples between the two systems without
exposure to the atmosphere. A schematic diagram of the combined electron
deposition/AES system is shown in Fig. 15.
ANALYSIS DE POSITION
Ion
Gun
ulato 1
l IonPump
SED
Gate
Valve
Ion
Gun
Crystal OpticatO O
Monitor Monitor
B
SampleE - BeamHearth I Transfer
Arm
Figure 15. Schematic diagram of the combined electron beam deposition andAuger analysis system use for facet coating experimentation.
The CSP samples used in this investigation were cleaved in atmosphere,
and no prior cleaning was employed prior to loading the samples into the
deposition system. In all cases the samples were examined prior to cleaning and
were subsequently examined at various stages during the argon ion-beam milling
process. The only impurities observed on the sample prior to ion beam milling
were carbon and oxygen. After milling, AES measurements revealed that the
carbon had been completely removed, while trace amounts of oxygen still
remained on the facet. The inability to completely remove oxygen is likely to be
due to absorption of residual water in the deposition system prior to the transfer
operation. Trace amounts of argon were also observed to be present after milling.
In Fig. 16, a plot of the normalized oxygen peak height vs the ion milling
time for 1000-eV argon ions at three different ion current densities is shown. The
rate at which the oxygen is removed from the facet is clearly proportional to the
27
ion current density, as was expected. Similar results have been obtained at ion
energies between 200 and 1000 eV. An estimate of 10 /_ for the native oxide
thickness was determined based upon the ion milling rates of A1203 and Ta205
standards. The oxide removal rates, using this thickness, were approximately 1,
2, and 5 /_/min. for ion current densities of 0.02, 0.03, and 0.04 mA/cm 2,
respectively. These results compare favorably with those observed by other
researchers [7]. A complete description of this work can be found in Appendix C.
50.0
"1-
.-< 40.0(D
13_
= 30.0(D
>..,X
o 20.0
(D
•_- 10.0
oZ
Figure 16.
0.0 ' ,
0.0 10.0
•-_ 0.04 mA/sq, cm.
.-c- 0.03 mA/sq, cm.
"-_ 0.02 mA/sq, cm.
20.0 30.0
Ion Sputtering Time (min.)
Removal rate of oxygen from the laser facet as a function of ion
sputtering time using 1000-eV argon ions.
D. LASER DIODE DIE AND WIRE MOUNTING
The problems associated with the use of indium solders are well known and
documented. Thus, we have focused our attention on the development of fluxless
mounting techniques incorporating hard solders. Some hard solders, such as
tin, are subject to "whisker" growth; therefore, we concentrated our efforts using
gold-based solder alloys (i.e., Au-Sn, Au-Ge, Au-Si, etc.). The soldering process is
performed in a hydrogen/nitrogen environment to inhibit the formation of oxides
during mounting. Analysis of devices on lifetest, using scanning electron
microscopy (SEM) and energy dispersive analysis of x-rays (EDAX) after
operating lifetimes in excess of 8,000 hours, have shown no evidence of the
"whisker-type" growth that had previously been seen using tin-based solders.
The wirebond connection to the n-side of the CSP laser has also been
examined. The first wirebond made to the CSP lasers was performed on a ball-
type wirebonding machine. The smallest-diameter gold wire (0.007 in.) having
28
the lowest hardness value was used to minimize the amount of stress placed on
the chip during the wirebonding process. The deformation of the ball and damage
to the laser chip were evaluated using SEM and metallurgy cross-sectioning
techniques. A series of experiments were conducted to fully assess the impact of
the stress on the laser chip during the wirebonding process. A 3-gm pull-force
value was used as the minimum acceptable bond strength. In almost all cases,
damage to the laser chip was observed, owing to the force required in deforming
the ball to obtain the minimum bond strength. Additional analysis revealed that
the exact cause of the damage was not only associated with the wirebonding
process but also with the poor quality of the wafer surface used for the wirebond.
During the thinning procedure, this surface is lapped to remove the residual zinc
diffusion and to reduce the wafer thickness for the subsequent cleaving process.
Analysis of the lapped surface revealed damage to the GaAs crystal at depths up
to 0.002 in. (total chip thickness after lapping is only 0.004 in.). During the
wirebonding process, the stress placed on the chip during the deformation of theball resulted in propagation of the defects associated with the lapping procedure
into the laser chip. A modification to the thinning process using the standard
lapping procedure in conjunction with a chemical etching procedure has not onlyreduced the damage associated with the lapping process but has dramatically
improved the surface finish. This has lead to wirebonds exhibiting less damage
with greater bond strengths. However, to ensure long-lived CSP lasers, it is
necessary to eliminate all damage associated with this bond. This requirement
leads us to investigate another bonding technique, called wedge bonding. This
technique still requires deformation of a wire, but not of the large diameter ball.
Thus, the forces necessary for deformation are greatly reduced. By utilizing our
experience with the ball bonding process and applying it to the wedge-style
bonding, we have been able to obtain bonds exhibiting the minimum bond strength
without any observable damage to the CSP laser chip. Analysis of failed-lifetest
devices after many thousands of hours of operation have shown no propagation ofdefects from the location of the wirebond to the laser structure.
In the CSP laser, the heat caused by operating the laser is generated along
the contact stripe on the p side of the chip. The hottest location along that stripe,
however, is near the emitting facets, where additional heat is generated due to
strong optical absorption. Thus, heatsinking of the facets, particularly the output
facet, is essential for reliable operation. The beam divergence of the CSP laser is
quite large, which requires that the position of the output facet of the chip be at the
29
edge of the heatsink. Mechanically polishing this edge to the tolerance required
for suitable heatsinking (radius <1 pm) is quite time-consuming and results in
the incorporation of the polishing media into the oxygen-free high-conductivity(OFHC) copper mount (measurements performed using AES techniques). Thus,
we have developed a broaching technique that leaves the corner of the OFHC
copper mount with a radius of less than 1 _m and without any contamination on
the mounting surface. In addition, the surface roughness or quality of the
mounting surface by both techniques is comparable.
E. LASER DIODE OPERATING CHARACTERISTICS
A CSP laser has been fabricated that has produced lasing operation to 190-
mW-cw, single-fundamental-spatial and spectral-mode operation up to 70 mW
cw, with single-spatial-mode operation continuing to 150 mW; beyond 70 mW
there are increasing line-broadening effects in the parallel far-field patterns
accompanied by the appearance and growth of spectral sidebands. We show in
Fig. 17 the power output vs current input (P-I) curves, the spectral content of the
output, and the parallel and perpendicular far-field radiation patterns at different
power levels. The laser facets for these measurements were coated with an
A1203/Si dielectric stack to produce 90% reflectance on the back facet and an
approximate %/4 A120 3 layer to produce an approximately 10% reflectance on the
front, or emitting, facet. The room temperature (23°C) cw threshold current is 48
mA and the differential quantum efficiency, Tl, at the emitting facet is 41%. The
laser displayed a minor kink in the P-I curve at =70 mW of output power. The
performance characteristics of the device remained unchanged over the entire
power range including the kinked region. However, broadening of the lateral far-
field radiation pattern, due to gain saturation and heating effects, could be
observed at powers >100 mW. The wavelength shift is that expected from the
bandgap shift due to joule heating and a 25°C/W mounted-diode thermal
resistance. The beam FWHP at 20 mW for the parallel and perpendicular far-
field patterns are, respectively, 6.5 ° and 27 ° . It is worth noting that, after failure
at 190 mW cw, the laser facet visually showed no damage and the laser continued
to be operable up to 120 mW cw.
3O
(Jv
/J
i/
t
tt
\
\
o
÷
WJ
Z
I
//
JJ
/
I Ii
\
+
UJl_9Z
A
A
0v
OO
-.J
(S.LIOA) A
(Mw) d
it)
o_
N
aD
o_0
Oit)N
E
O
(S.L"IOA) A
_r
I
0
(/_,w) d
O_3
c%)
_<1[
¢%)
O
E
O
L_
31
0_IN
//
//
/.,......I
J
oI
+
W
IN
(s_q0^) ^
0
(MW)d
N
E
0
//
//
t
c%)
I0
($1qOA) A
(Mus) d
0
+
I,iJ.Jc_Z
0
I
q.cO
0v')
cO
8
4(E
0
oF.q
0
_S
• ,,..q
f..,
v 0
•._ :::::S
0 _
g
32
i
m
m
m
m
I I I I I
2 nsec/div
Current input pulse10% duty cycle
I I
2 nsec/div
Peak pulse opticaloutput power = 60 mW
m
i
I
m
B
m
m
I I I I I I I I I
2 nsec/div
Peak pulse opticaloutput power = 80 mW
Figure 18. Response of a CSP high-power laser to square current pulses at 14%duty cycle. The fall and rise times are <0.5 ns.
33
Typical laser modulation behavior is indicated in Fig.18, in which we show
the laser response to square current pulses at 14% duty cycle. The fall and rise
times are <0.5 ns (the limit of the pulse resolution); note the almost complete
absence of tailing in both the leading and trailing edges of the output pulse, as
well as a minimum of ringing oscillation. Modulation properties were found not
to change at power levels up to 80 mW, the limit of the experiment.
F. LIFETESTING AND RELIABILITY ASSURANCE
Semiconductor lasers, such as the CSP lasers, only operate in the lasing
mode at rated Power up to about 100°C. Thus, there is little margin above the
normal maximum operating temperature to carry out fully operational lifetests
that can be used to obtain traditional Arrhenius extrapolations. As a
consequence, we use detailed measurements of the change in drive current for
rated power at operating temperatures of 25°C, 50°C, and 70°C. Previous studies
[12] have shown that the change in threshold current and drive current could be
described by a power law of the type
T = Atn
where 0 < n < 1. By fitting this expression to the characteristics of the aging
parameter, the predicted change can be determined. All the CSP lasers were
tested using a constant power technique in which the drive current used to power
the device is continually adjusted to maintain rated output power. This technique
subjects the devices to a greater operational stress than the constant current
technique in which the laser is placed on lifetest at rated power and the decrease
in output power is monitored.
A total of 15 CSP lasers were placed on lifetest; 8, 4, and 3 at operating
temperatures of 25°C, 50°C, and 70°C, respectively. In all cases the operating
conditions were 50 mW at a 50% duty cycle and at a repetition rate of 10 MHz. The
threshold currents for all the devices were below 80 mA, and the initial thermal
resistances were between 20°C/W and 40°C/W. The lasers were mounted p-side
down on a copper heatsink and had (k/4) A1203 coating with a reflectance of 10%
on the emitting facet and a (_J4) AI203/Si stack coating having a reflectance of 90%
on the rear facet. The devices all operated in the fundamental mode and had
emission wavelengths between 8600 and 8800/_. The devices were not subject to a
34
preselection burn-in process prior to placement on lifetest. The devices placed on
lifetest were taken from three different LPE growth runs that displayed
operational characteristics of suitable quality for use in the ACTS program.
As shown in Fig. 19, all eight devices placed on lifetest at 25°C are
continuing to lase after 1000 to 1500 h on lifetest. In most cases only a modest
increase in drive current was required to maintain the rated power of 50 mW.
CSP lasers removed during the lifetesting process and re-characterized exhibited
no change in the far-field radiation patterns or spectral characteristics, and only
a small change was observed in the threshold current value. The invariance in
the far-field pattern means that the lasing spot remains unchanged.
In Fig. 20, we show the results of four CSP lasers placed on lifetest at 50°C
at 50 mW. In addition, we have included a projected lifetime for the lasers if an
activation energy of 0.07 eV is assumed. Two of the devices have not required any
change in drive current after 500 and 800 h, respectively. The other two devices
have required large changes in drive current, which would be unacceptable in the
ACTS program. Although these devices are considered unacceptable, the
projected lifetimes of the devices still exceed 6,000 h, more than 3 times the
estimated lifetime expected at the beginning of this program.
500 -
2_400
E
Z
hln"
rr 30CD
0
u)<
2oo
I00
Figure 19.
pO = 5omW
860-880nm
(DUTY CYCLE:50% ,IOMH;" )
ol , I l IO tOOO 2000
TIME(HRS)
Aging behavior of high-power, 8600- to 8800-/k CSP lasers at 25°C as a
function of operating time. The lasers were maintained at a constantoutput power level of 50 mW (50% duty cycle; 10 MHz).
35
5OO
400
[300
0
200
IO0 f
00
D
PC: 50mW
860-880nm
{ DUTY CYCLE= 50% , IOMH;_ )
I L I L I L I I I
200 400 6,OO 8OO IOOO
ACTUAL TIME (HRS)
I I I I I I I I I
O 2,000 4,000 6,000 8,000PROJECTED LIFETIME (HRS)
Figure 20. Aging behavior of high-power, 8600- to 8800-A CSP lasers at 50°C as a
function of operating time. The lasers were maintained at a constantoutput power level of 50 mW (50% duty cycle; 10 MHz).
The results of the three devices placed on lifetest at 70°C and 50 mW are
shown in Fig. 21. As before, we have included a projected lifetime scale for these
devices, assuming an appropriate activation energy. One of the devices placed on
lifetest started to degrade immediately and most probably should not have been
used. However, even with this rapid degradation, its usable life was in excess of
5,000 h, as predicated from our projected lifetime value. The other two devices
degraded in a similar manner and are still operating after 450 h on lifetest.
Although these lifetests are continuing, if we assumed they were completed at
this point, the projected lifetimes for these devices would exceed 17,500 h, a value
that would be acceptable for ACTS-type devices. This value translates to
approximately two years of usable life if the devices were used on a continuous
basis. If the devices were used intermittently, their lifetimes would be
commensurate to the time they were in operation.
36
5O0
400
E
F-Z 300Idrr
30
200<
I00
Pc= 50 mW
86o-s8onrn
(DUTY CYCLE:50% . IO MHE )
o i I i I J I i I IO IOO 200 300 4.00 500
ACTUAL TIME ( HRS )
I ' I _ I ' f ' IO 5,000 IO,OOO 15,OOO 20.000
PROJECTED LIFETIME (HRS)
Figure 21. Aging behavior of high-power, 8600- to 8800-/_ CSP lasers at 70°C as afunction of operating time. The lasers were maintained at a constantoutput power level of 50 mW (50% duty cycle; 10 MHz).
G. POST-LIFE FAILURE ANALYSIS
The examination of failed devices is critical to the development of new
techniques and processes that will ultimately lead to CSP lasers that exhibit high
reliability and long life. The analysis of a failed device is not a straightforward
process, owing to the inherent small size of the laser chip and the even smaller
geometry of the laser structure. Thus, many of the routine analytical techniques
had to be modified so that the analysis on the small laser chip could be performed.
The diffusion of zinc into the CSP structure is an important part of the
fabrication procedure. It provides the high carrier concentration necessary for
good ohmic contact and confines the current to the lasing region of the device. In
addition, the position of the zinc front determines the gain profile for the laser.
Analysis of devices on lifetest revealed that some of the devices from different LPE
growth runs whose near-field radiation pattern prior to lifetesting displayed a
single lobe that had changed to a two-lobe far-field after failure in lifetesting. In
Fig. 22(a) and (b) the near-field radiation patterns for a CSP laser before and aider
37
lifetest are shown. We refer to the reduced intensity in the center of the near-field
pattern of Fig. 22(b) as a node. The devices that exhibited these effects were
examined using angle lapping techniques in an attempt to identify the common
physical feature that would account for the observed effect.
In almost all the devices we examined with this observe defect, the common
feature was the position of the zinc front in the structure to that of the active layer.
In some cases the zinc front that we delineated using a chemical etchant had
actually penetrated into the active layer, while other devices displayed a front in
very close proximity to the active layer. The deep zinc diffusion region is not
composed of a uniform concentration of zinc. Analysis of this region by secondary
ion mass spectroscopy (SIMS) has shown that the leading edge of the zinc region
may be one to two orders of magnitude lower in carrier concentration than the
surface. Thus, delineation by chemical etching techniques may not reveal the
actual position of the zinc front, but instead identify a specific carrier
concentration at which the etching process is activated. However, SIMS studies
have shown that the actual misplacement between the zinc and the etched fronts
is <1000/_. Thus, for our analyses here we can assume that both fronts are at the
same location. The evidence linking the position of the zinc front to a change in
the near-field pattern suggests that the two are related.
Previous work by other researchers [8-10] suggests that a highly doped,
zinc-diffused region is a potential source of defects that form non-radiative
recombination centers that degrade the lifetimes as well as the performance of
laser diodes. A nonradiative recombination region in the active layer of the device
will result in no light being emitted from the area. Thus, devices with the zinc
front penetrating into the active layer will display near-field radiation patterns
containing two lobes, each lobe corresponding to a lasing spot on either size of the
nonradiative region. A process change has been implemented to eliminate this
problem in the future. This change involves growing a thicker p-cladding layer
(from 1 to 1.5 _m) in the CSP structure and maintaining the zinc front, as
delineated by a chemical etchant, at least 0.5 _m from the active layer. Devices
meeting these critical parameters have been placed on lifetest, and no change in
the near-field radiation pattern has been observed after many thousands of hours
on lifetest. The complete details of this study can be found in Appendix E.
38
5 I_m
(a)
5_a
Figure 22. (a) Pre-lifetest, near-field pattern and light-intensity scan for high-
power CSP laser. (b) Post-lifetest, near-field pattern and light-intensity scan for the same CSP laser.
ORIGINAL PAGE TS
OF POOR QUALITY
39
III. LASER PERFORMANCE OF A DFB-CSP LASER
The goal of this phase of the program was the development of a high-power
CSP laser that operates in a stable, single-longitudinal mode. Our approach to
achieving this single-longitudinal-mode behavior is the introduction into the CSP
structure of an additional feature rather than merely relying on the
reproducibility of those elements that produce the laser's wave-guiding
properties. To achieve stable operation at a single wavelength, a grating will be
incorporated directly into the CSP structure, making it a distributed-feedback,
channeled-substrate planar (DFB-CSP) laser. This grating, whose geometrical
properties are determined from the laser's planned wavelength and refractive
index, can be built directly into the CSP structure. The periodic grating instead of
the mirror facets will now provide the feedback to support lasing. The feedback
via the grating should produce a device with stable, single-longitudinal-mode
operation even under 100% depth modulation conditions as well as providing
improved temperature dependence of the longitudinal mode (0.7 /_J°C for DFB-
CSP vs 3/_°C for standard CSP). In addition, since the feedback for lasing is
provided by the grating, this type of structure should be much less susceptible to
instabilities in the longitudinal-mode behavior, owing to light being reflected back
into the lasing cavity from various components (i.e., lenses, beam splitters, fibers,
etc.) in the optical system.
A. DEVICE MODELING
In conventional lasers, the optical feedback is provided from a pair of
reflecting surfaces that form a Fabry-Perot cavity. In a DFB laser, optical
feedback is provided from a Bragg-type diffraction grating. In DFB lasers the
grating is usually produced by corrugating the interface between two of the
semiconductor layers that comprise the laser. This corrugation provides 180 °
reflection at certain specific wavelengths, depending on the grating spacing.
The basis for selective reflection of certain wavelengths can be understood
by examining the original case considered by Bragg [11], in this case, the
reflection of an incident plane wave by the atomic planes of a crystalline lattice.
However, the same effect is observed in the case of reflection from a corrugated
grating formed in the junction plane of a laser. To maintain the phase coherence
of the plane wavefront and thereby avoid destructive interference, the path lengths
40
for reflections from successive reflectors must differ by an integral number of full
wavelengths. Thus, from geometrical considerations, we obtain the Bragg
relation given by
2d sin0 = _., _ = 1, 2, 3,...,
where 0 is the angle formed by the incident ray and the atomic planes, and _. is
the optical wavelength in the medium. To adapt this relationship to the case of
180 ° reflection by a grating in a DFB laser, it is only necessary to let d equal the
grating spacing A, let _. equal )_0/ng, where ng is the effective index in the
waveguide for the mode under consideration, and let 0 equal 90 °. Under these
assumptions the equation above becomes
2 A= _(_.0/ng), Z = 1, 2, 3, ....
The vacuum wavelength of light that will be reflected through 180 ° by such a
grating is therefore
ko = 2Ang/A, _ = 1, 2, 3, ....
Although the grating is capable of reflecting many different longitudinal
modes, corresponding to the various values of _, usually only one mode will lie
within the gain bandwidth of the laser. In fact, because of the difficulty of
fabricating a first-order (i.e., A-- 1, A =1200 /_) grating, usually a second-order
grating is used in most DFB-type structures.
The change in lasing wavelength as a function of heatsink temperature for
a DFB laser can be calculated if the principal of operation is understood. The
wavelength shift in a DFB laser results from a change in the Bragg condition due
to changes in both the index of refraction and in the grating period. Since the
fractional change in index with temperature is about an order of magnitude
greater than the thermal expansion coefficient, only the former need be
considered.
In a DFB laser, the lasing wavelength is locked by the grating's Bragg
wavelength, given by
_-B = 2nA/_
41
where n is the effective index, and • and A are the grating order and period,
respectively. Differentiating this equation with respect to temperature yields
dT - ng _.B
where ng, the "group" index is given by
and dA/dT is assumed small. For an A1GaAs laser operating at 8300/_, ng is =4.2
(from the Fabry-Perot mode spacing). Thus, for
XB = 8300/_ and bn/_}T = 3.5 x 10 -4 °C-1
we find that the wavelength change with temperature should be about 0.7/_/°C.
In addition to providing a means of accurately selecting the peak emission
wavelength, grating feedback also results in a narrower linewidth of the optical
emission. The spectral width of the emission line is established by a convolution
of the laser gain curve with the mode-selective characteristics of the laser cavity.
Since the grating is much more wavelength-selective than a cleaved facet, the
resulting emission linewidth of a DFB laser is significantly less than that of
Fabry-Perot laser. The emission linewidth of a DFB laser depends strongly on the
coupling coefficient k of the grating. Decreasing the operating temperature also
decreases emission linewidth, because the gain curve of the laser is narrowed as
the thermal distribution of electron energies is decreased.
B. GRATING FABRICATION AND LPE/MOCVD GROWTH TECHNIQUES
A schematic of the structure is shown in Fig. 23(a) and in the stained cross-
sectional cleave in Fig. 23(b). The details of the structure can be seen more clearly
in Fig. 23(c), in which we show a stained cross-section lapped at a 1 ° angle in the
vertical direction. The substrate is prepared by chemically etching a second-order
grating using a photoresist mask obtained by standard holographic
interferometry into a 0.84tm-thick, n-type Alo.15Gao.ssAs layer. A 0.124tm-thick
42
GaAs layer is then grown over the grating. These layers are grown using
metalorganic, chemical-vapor deposition (MOCVD) for improved uniformity and
surface morphology. The MOCVD growth also improves the nucleation of the
GaAs layer on the Alo.15Gao.ssAs grating surface. The Alo.lsGao.8sAs layer
prevents layer meltback of both the V-channel and the grating, while the GaAs
layer grown over the grating, provides a nucleating surface during subsequent
LPE growth. In Fig. 23(c) some nucleation problems are still evident with the
MOCVD grown GaAs layer. The index step at the grating interface is
approximately 0.1.
DFB-CSP LASER
CROSS-SECTIONCLEAVE
S,O 2
CAPP-CLAO
/ ACTIVEN-CLAD
J_- PROT EC TIVE
-- GRATING
ANTq- MELTBACK
SUBSTRATE
(a)
CROSS-SEC_ON
1°ANGLE-LAP
n-GaAs cap
p-clad AIGaAs ,LPE
n-active AIGaAsn-clad AIGaAs
n-buffer GaAs 1
Gratingn-AIGaAs MOCVD
(anti-meltback)
n-GaAs substrate
(c)
(b)
Figure 23. (a) Schematic diagram of CSP-DFB laser. (b) Stained cross-sectionalcleave of CSP-DFB structure. (c) Stained cross-sectional cleave lapped
at a 1° angle in the vertical direction. Note especially the beginning of
meltback between the n-cladding and n-buffer layers.
The LPE growth is performed in the automated LPE growth reactor, as
described in the LPE growth of CSP lasers section of this report. The MOCVD was
performed in Cambridge Instruments' MR-100 system at atmospheric pressure.
It contains four metal alkyls and three hydride sources for the growth of AIGaAs
compounds. The gas-handling system has been designed to give abrupt interfaces
(<10 /_) so that quantum-well-type structures can be routinely grown. Under
computer control, the gas flows and growth temperature may be ramped, making
ORIGINAL PAGE" IS
OF POOR QUALITY
43 ORIGINAL _ =;"_"
OF POOR QUALITY
it possible to grow graded-interface structures and devices that require graded
doping profiles as well. The horizontal cell configuration allows the thickness
and compositional uniformity over a 2-in. wafer to be better than 10%. The typical
background doping and mobilities in GaAs are -1014 cm -3 and 7500 cm2/V-s at 300
K and 80,000 cm2/V-s at 77 K. A photograph of the system is shown in Fig. 24.
Figure 24. Photograph of metalorganic chemical vapor deposition (MOCVD)system used for the epitaxia| growth of the DFB-CSP laser.
After the grating, which is formed using chemical etching techniques, and
its protective layers have been fabricated, a 4.2-pm-wide V-channel is chemically
etched into the substrate and four layers are grown: (1) Al.33Ga.67As cladding
layer (0.3 _m, Nd = 1 x 10 ls cm-3), (2) Alo.o6Gao.94As active layer (0.07 _m, Nd = 1
xl017 cm-3), (3) Alo.33Gao.67As cladding layer (1.5 _m, Na = 5 x 1017 cm-3), 4) GaAs
capping layer (0.7 _m, Nd = 5 x 1017 cm-3). The growth is performed at 800°C,
using a cooling rate of I°C. This procedure significantly reduces the complexity of
fabricating the DFB-CSP structure. In addition, the current path in the device
does not include a regrowth interface, which has been associated with higher-
than-normal series resistance for devices fabricated with one or more growth
44
interfaces in the current flow path. After LPE growth, routine contacting
technology completes the device, which includes Zn stripe diffusion for current-
confinement, Ti/Pt/Au for the p-contact, and AuGe/Ni/Au for the n-contact.
Upon closer inspection of fabricated structures, we have found that the
chemically etched gratings used in our structure were relatively shallow (-100/_)
and displayed non-uniformities across the surface of the wafer. Both the shape
and the depth of the grating should be considerably improved by the use of our
newly acquired Commonwealth Scientific ion-beam etching system. This system
can be operated in three different modes: first, conventional ion-beam etching,
second, reactive ion-beam etching, and third, ion-beam-assisted etching.
Conventional ion-beam etching isotropically removes material from a sample as a
result of bombarding it with inert gas ions. In the reactive ion-beam etching
mode, a reactive gas is used in place of the inert gas to add a chemical etching
component to the process. Ion-beam-assisted etching uses an inert gas ion beam
while a "blanket" of reactive gas is maintained at the sample surface. In this
case, the sample is not only chemically etched, but it is also protected from high-
energy ions that can cause crystalline damage.
The improvement in the quality of the grating using ion-beam etching
can be clearly seen from Figs. 25 and 26. In Fig. 25, a cross-sectional scanning
electron microgragh (SEM) of a grating prepared using chemical etching
techniques is shown. The grating is etched along the dovetail direction in the
GaAs substrate since the CSP structure is fabricated along the V-direction. Thus,
shallow gratings (200 to 300/_) are obtained, owing to the undercutting action of
the preferential etchant. This undercutting limits our ability to fabricate deep
gratings, which are required to enhance feedback and suppress Fabry-Perot
operation. In addition, a more complex photolithographic process is required for
chemically prepared gratings as opposed to ion-beam-etched gratings. This
photolithographic process leads to nonuniform linewidths in the submicron
grating pattern. These problems have led to the development of a grating
procedure incorporating ion-beam etching techniques.
A cross-sectional micrograph of an ion-beam-etched grating in a GaAs
substrate is shown in Fig. 26. The increased grating depth and sharply defined
mesas can be clearly seen in the photograph. The grating was fabricated using a
beam voltage of 300 eV and a current density of 0.7 mA/cm 2 at an incidence angle
of 30 ° .
45
Figure 25. A grating with a 2400-_ period formed in an AIGaAs layer by chemicaletching techniques.
._m x 30 0_O 30kV
Figure 26. A grating with a 2400-/_ period formed in an AlGaAs by ion-beammilling techniques.
46
Although we have been able to fabricate DFB-CSP lasers using the
techniques described above, we are still plagued with nucleation non-uniformities
and low yield. As a consequence, a series of wafers were grown with ion-beam-
milled gratings on the shoulders of the V-channel. The grating was fabricated
into a 0.6-pm-thick Alo.12Ga0.ssAs, layer and a GaAs layer was grown over the
grating using MOCVD techniques. The period of the grating was chosen based
upon emission wavelength spread obtained from previous LPE growths. All the
wafers grown displayed some non-uniform nucleation across the growth surface.
The growth in the channel region appeared normal, while the surrounding areas
showed various stages of growth. Analysis using angle-lapping techniques
revealed a problem at the interface between the grating and the GaAs protective
overgrowth layer. Further investigation on the LPE substrates identified that the
GaAs layer was non-contiguous.
A growth study was then conducted to determine the cause of the non-
contiguous film. A set of wafers containing gratings fabricated into a MOCVD
grown Alo.12Gao.ssAs layer were prepared for this study. Very thin GaAs layers
(100 to 1000/_) were grown on top of the grating and cross-sectional samples were
examined using a SEM. The samples showed that unlike conventional MOCVD
growth, the re-growth layer does not necessarily follow the previous surface
morphology in all situations. Figure 27(a) shows an SEM cross-sectional
photograph of an ion-beam-milled grating in a GaAs substrate. The grating has
a period of 2400/_, a value comparable with the period in our device structures,
and a depth of 1200/_. The slight waviness at the edges of the grating is an
artifact of the SEM used to take the photographs. Figures 27(b) and 27(c) show the
same grating milled into 14% and 23% AlAs but with A1GaAs and GaAs grown
over top of the grating. For the case of a thin Alo.4oGao.6oAs layer grown on top of
the grating (Fig. 27b), the layer follows the periodic nature of the grating
replicating its shape. On the other hand, when a thin GaAs layer is grown over
top of the grating milled in Alo.23Gao.77As (Fig. 27c), the layer fills in only the
valleys of the gratings. This results in a surface composed of GaAs and
Alo.23Gao.77As. This type of surface makes the growth of a layer on top by LPE
very difficult since the nucleation process will normally be incomplete across a
large area due to the exposed Alo.23Gao.77As regions. Thus, for obtaining
complete nucleation of the surface by LPE growth techniques, it is necessary that
the valleys in the grating be filled until coalescence of the GaAs material in the
I=Is
l_.
mmi Illll
E
I= o,==:O
(I,)
Illllnl V
r,#)
o
o
on
==:
_0I=
Jln
=._
==1:
oo
,=;i
,,I-,*¢'t3L...
¢,=
4?
v
ORiGiNAL ,:,,,',c _,SOF POOR QUALI'I'3f
el:
Im
L_
(4
N
m
L.
r,#}
IE=I.
__2_
as_
° _
,.., _ es
g
48
valleys occurs above the peaks of the grating. This coalescence will ensure
uniform nucleation during the subsequent LPE growth process.
C. LASER DIODE OPERATING CHARACTERISTICS
Theoretically, successful DFB action should be indicated by two lines in the
output spectrum. In practice, however, only one line is most often seen because of
non-uniformities or asymmetries in the Bragg scattering or reflections in the
Fabry-Perot cavity. Thus, the significant indication of DFB action is that the
spectrum of light output contains one line, indicating a single longitudinal mode,
that remains essentially fixed with no mode-hopping over a range of power and
temperature for both cw and pulsed operation. This wavelength can, in fact,
change with operating conditions, such as power and temperature, but it changes
largely according to temperature-induced changes in the effective index, which
can be predicted theoretically (see modeling section). The mode remains locked by
the grating rather than determined by the properties of the Fabry-Perot cavity,
and changes observed are those stemming from the grating.
8320
o<
=.
• 8310
' I ' 1 ' I ' I ' J '
Figure 28.
m
>m
Temperature Shift forDFB-CSP Laser
8300 i [ I I I I ; I I I I I (lOOnsec' IOSHz)22 24 26 28 30 32 34
Temperature (°C)
Emission wavelength shift as a function of heatsink temperature for aCSP-DFB laser operating pulsed at an output power of 10 mW.
The CSP-DFB laser displayed DFB operation at pulsed (1% duty cycle; 100
ns) power levels up to 40 mW and at cw power levels up to 10 mW. Figure 28
shows the change in the single-longitudinal-mode spectrum for a DFB-CSP laser
operating over the temperature range of 24°C-32°C. The pulsed output power from
49
the device was maintained at 10 mW. The facets of the device were coated (front-
10%; rear-85%), and no attempt was made to suppress the Fabry Perot modes.
Sideband rejection ratios are in the range of 18 to 23 dB. The observed pulsed
spectra were characteristic of DFB lasers in that they remained single-line for all
temperatures [6]. A typical pulsed spectrum for this device operating at 28°C can
be seen in Fig. 28. The wavelength temperature dependence for both pulsed and
cw operation is about 0.7/_/°C. This behavior is that expected for an A1GaAs laser
of the given composition and layer thickness operating at 8300/_ [7]. The fact that
the DFB behavior described above occurs over a relatively small temperature
range and at powers only up to 40 mW pulsed can be associated with the small
grating depth (-100 to 200 /_) produced by chemical etching in the dovetail
direction.
40
E
W
0E
P-I-V CURVE FAR-FIELD PATTERN
I I I I I I I I i I , , _ I ' ' ' '
-- CW
--Vl
...... 50
i I
/I/
/III
/;III
I
/
/o _ P i I i i , I j
0 I O0 _'00
CURRENT (mA)
_tl ,i.I I I
----.------ Poro Cw-IO _ 'i Ii
..... Pore CW'20 I I l...... Perl_ CW-203 I
II I I
i II!I
I
!I
I
'jI I
/I J
ss/// l
C) L_ _L_ I--50
I
I i
i I
i II
tt
tI I"_. I I I
+5(
ANGLE, DEGREES
(a) (b)
Figure 29. (a) Power-current curves for a CSP-DFB laser. (b) Far-field radiation
patterns for a CSP-DFB laser.
Figures 29(a) and (b) show the P-I curve and far-field patterns, respectively,
demonstrating thresholds as low as 50 mA, kink-free power curves, and well-
defined, single-spatial modes. The overall efficiency of the laser at 40 mW (total
input electrical power divided by output optical power) is 15%. Thus,
incorporation of the grating and the accompanying extra layers (see Fig. 23) do
50
not interfere with the desirable spatial mode and high-power properties of the
basic CSP structure. A more complete description of the results can be found in
Appendix F.
51
IV. CONCLUSIONS
The continued support of the high-power CSP laser by NASA has led to
significant improvements in both the performance levels and reliability of the
device at wavelengths (8600 to 8800/_) previously thought unattainable. These
results are important because the transparency of the atmosphere in this
wavelength regime makes possible optical communication links for space-to-
ground systems. Advances in LPE growth and computer modeling techniques
has resulted in the fabrication of lasers operating at record high-power levels. In
addition, these device improvements have led to overall improvements in
reliability both in terms of output power and operating lifetimes. Further work is
still necessary, however, to develop laser sources to meet the critical requirements
for space applications. Both higher power capability and reliability improvements
are necessary for the realization of a spaceborne laser communication system.
As A1GaAs lasers continued to be used in a wider range of applications, it
has become increasingly apparent that control of the longitudinal mode will be
necessary to avoid mode-hopping and instabilities in the output of the laser due to
optical feedback. The development of the DFB-CSP lasers addresses these issues.
This novel structure has demonstrated stable longitudinal-mode operation over
an 8°C temperature range at pulsed powers up to 40 mW. However, if a laser
device is to be incorporated into space-qualified systems, more work needs to be
performed. Some of the specific areas that need to be explored are the
incorporation of the grating in the peak of the optical mode, improved grating
fabrication, and the development of new growth techniques to extend the usable
power and temperature range.
53
REFERENCES
. D. B. Carlin, M. Ettenberg, N. A. Dinkel, and J. K. Butler, "A high-power
channeled-substrate-planar A1GaAs laser," Appl. Phys. 47, 655-667 (1
October 1985).
. G. A. Evans, J. K. Butler, and V. Masin, "Lateral optical confinement of
A1GaAs channeled-substrate-planar lasers," IEEE Journal of Quantum
Electronics, QE-24 (5), 737-749 (May 1988).
. J. K. Butler, G. A. Evans, and B. Goldstein, "Analysis and performance of
channeled-substrate-planar double-heterostructure lasers with geometrical
asymmetries," IEEE J. Quantum Electron, 0E-23 (11), 1890-1899 (November
1987).
. J. J. Hsieh, "Thickness and surface morphology of GaAs LPE layers grown
by supercooling, step-cooling, equilibrium-cooling, and two-phase solution
techniques," J. Crystal Growth 27, 49 (1974).
o H. Kressel and J. K. Butler, Semiconductors Lasers
LED_, Chapter 7, Section 5 (Academic Press,
pp. 230-234.
and Heteroiunction
New York, 1977),
o K. Aiki, M. Nakamura, T. Kuroda, J. Umeda, R. Ito, N. Chinone, and
M. Maeda, "Transverse mode stabilized AlxGal_xAs injection lasers with
channeled-substrate-planar structure," IEEE J. of Quantum Electron.,
(2), 89-94 (Feb. 1978).
° S. P. Kowalczyk, J. R. Waldrop, and R. W. Grant, "Interfacial chemical
reactivity of metal contacts with thin native oxides of GaAs," J. Vac. Sci.
Technol., 19 (3), 611 (1981).
. I. Ladany and H. Kressel, "Influence of device fabrication parameters on
gradual degradation of A1GaAs cw laser diodes," Appl. Phys. Lett. 25 (12),
708 (15 December 1974).
PRECEDING PAGE BLANK NOT I_ILMED
54
9. I. Ladany and H. Kressel, "GaAs and related compounds, 1974," Inst. Phys.
and Phys. Soc., Conf. Ser. No. 24, London, pp. 192, 1975.
10. H. Kressel and J. K. Butler, Semiconductors Lasers and Heteroiunction
LEDs, Chapter 7, Section 5 (Academic Press, New York, 1977), pp. 543.
11. R. G. Hunsperger, Intearrated Optics Theory and Technology, Chapter 13
(Springer-Verlag, New York, 1985), pp 215.
12. D. Botez, J. C. Connolly, M. Ettenberg, D. G. Gilbert, and J. J. Hughes,
"Reliability of constricted double-heterojunction A1GaAs diode lasers," Appl.
Phys. Lett., 43 (2) (15 July 1983).
Appendix A
A
57
SELF-CONSISTENT ANALYSIS OF GAIN SATURATION
IN CHANNELED-SUBSTRATE-PLANAR
DOUBLE-HETEROJUNCTION LASERS*
by
Jerome K. Butler** and Gary A. EvansDavid Sarnoff Research Center
CN 53O0
Princeton, NJ 08543-5300
ABSTRACT
A self-consistent model for semiconductor lasers (using the CSP-DH laser structure
as an example) which does not assume constant opt/ca/power along the laser axis is
developed. This approach allows for the analysis of high power lasers with low facet
r_flecfivifies which produce nonuniform photon densities along the propagation direction.
Analytical equations for the modal gain coefficient, the threshold current density, and (for
facet reflectivities > 0.2) the radiated power for a specific CSP laser structure are obtained.
L
This work was supported in part by NASA, Langley Research Center, Hampton, Virginia under ContractNumber NAS1-17441.
"'Souflaem Me_odist University, Dallas, "IX 75275.
PRI_CI_DtNG PAGE BLANK NOT FILMED
59
A SELF-CONSISTENT ANALYSIS OF' GAIN SATURATION
IN CHANNELED-SUBSTRATE-PLANAR
DOUBLE-HETEROJUNCTION LASERS*
by
Jerome K. Butler** and Gary A. EvansDavid Sarnoff Research Center
CN 5300
Princeton, NJ 08543-5300
Semiconductor lasers with high reflectivity facets (> 0.3) have an almost uniform
density of photons along their longitudinal axis, allowing the assumption of uniform gain
in their analysis. However, the highest power from a semiconductor laser is achieved by
using a high reflectivity end facet and a low reflectivity (0.05 to 0.1) output facet resulting
in large variations of the photon density along the cavity axis 1-4. In this paper we extend a
recent uniform (longitudinal) gain analysis 5 based on the "self-consistent model". This
extension, which allows for photon density variations along the longitudinal axis, results in
relatively simple phenomenological equations for the modal gain coefficient, the threshold
current density, and (in the limit of high facet reflectivities) the radiated power.
In self-consistent models the carrier density in the active layer is derived from a
solution of the diffusion equation having both source and sink terms. The source for the
injected carriers is the drive current, whereas the sink is the stimulated recombination. The
current flow into the active layer varies laterally and lateral carrier diffusion within the
active region affects the optical gain profile modifying the shape of the optical field
distribution in the laser diode.6-14 The spatial dependence of the recombination term is
computed from the product of the lateral gain profile and the photon density in the active
layer as a function of position along the longitudinal axis.
*This work was supported in part by NASA, Langley Resea_h Center, Hampton, Virginia under ContractNumber NAS 1-17441.
**Southern Methodist University, Dallas, TX 75275.
60
For wave propagation of the form exp(jc0t - )'z), the transverse electric field
polarized along y can be solved using the effective index method; it is written as
Ey= E0 u(x,y) v(y) exp(-'?'z) (1)
where the complex function u(x,y) determines the transverse field shape along x, but it is
slowly varying (along y) compared to v(y) which defines the lateral field profile. The
nonlinear differential equation for the lateral field v(y) satisfies
d2v
d___+ [?2 _ y02 + k02F¢(y) Ka(y,Va)] v = 0 (2)
where Fc(y) is the complex confinement factor and k0 is the free-space wavenumber. The
complex effective index of refraction is neff = -j3'0(y)/k0 and 70(Y) is found from the
solution of the transverse problem. The value Ka(Y,Pa) is the carrier dependent part of the
dielectric constant in the active layer and Pa is the optical power in the active layer. In terms
of the active layer gain g(Y,Pa), the dielectric perturbation can be expressed as
Ka(Y,Pa) = na g(y,Pa)(2R + j)/k0 (3)
where na is the passive index of the active layer and R is the gain induced index
suppression coefficient.
The carrier distribution in the active layer must be found from solutions of the
inhomogeneous diffusion equation having both "source" mid "sink" terms. The source
term is Rpump = Jx(Y)/qda, where :Ix(y) is the current density, q is the electronic charge,
and da is the active layer thickness. For typical stripe geometry laser structures with Zn
diffusion fingers, the lateral variation of the current density can be fit to the simple analytic
formula 15
Jo ;lyl < S/2
Jx(Y) = Jo (4)
[l+(lyl - S/2)/yo] 2 ;lyl > S/2
where Jo is the current density under the stripe; YOis primarily a function of the Zn
diffusion depth, the stripe contact width, and the resistivities of the various epitaxially
61
grownlayers.Further,the injectedcurrent density is assumed to be invariant with z. The
sink term is Rst = P F(y) Ivl2 g(Y,Pa) where P is the intracavity power, F(y) is the intensity
confinement factor of the active layer, and Iv[2 is normalized to unity over (.oo,oo).
The diffusion equation governing the pair density N(y) is
d2N NDe B N 2 = -Rpump + Rst (5)
dY2 'Cs
where De is the effective diffusion coefficient, Xs is the carrier lifetime, and B is the
bimolecular recombination coefficient. The gain coefficient is g(Y,Pa) = aN(y) - b where
a = 2.5 x 10 -16 cm 2 and b = 190 cm -1. In general, the stimulated recombination term Rst
will be functionally dependent on z. However, we have neglected carrier diffusion along
the longitudinal direction since I_N/3zl << N/LD, where LD is the carrier diffusion length.
At a given point z, the intracavity power P(z) = P+(z) + P-(z) is the sum of the
forward and backward waves. The power in the active layer Pa is computed from a
fraction of the intracavity power P. In lasers with high facet reflectivities, the intmcavity
power is almost constant along the z direction, however, when reflectivities are small, P(z)
has relativity large changes along the axis. The modal power gain coefficient
G = -2 Re{7} is computed from the self-consistent solutions of the carrier diffusion and
Maxwelrs equations.
Solutions of the optical fields of the CSP-DH structure with the material parameters
shown in Fig. 1 are obtained with an index supression coefficient R = -2. The parameters
used in the diffusion equation are LD = D_e_s = 3 p.m, Xs = 3 ns, B = 10 -10 cm3/s, the
stripe width S = 6 p.m, and Y0 = 0.5 p.m. The points in Fig. 2 show numerically computed
values of the modal gain coefficient as a function of the intmcavity power for different
values of drive current which has been normalized for a laser length L = 100 p.m.
62
Although thenumericaldatacanbeusedto calculatethelongitudinalvariationof the
gain in a laserwith knownfacetreflectivities,it isusefulto fit thenumericallycalculated
modalgaincoefficientpointstoananalyticalexpression(motivatedby theform of the
equationfor gainsaturationin homogeneouslyandinhomogeneouslybroadened
systems)16writtenas
G(P)= (_) c GO(1 + P/Ps)d oq (6)
wherec, d, GO,Ps,andal areunknownconstants.An optimizationprocedurefor a least-
squaresfit of (6) to thecomputedvaluesof theself-consistantmodelgivesc = 0.708,
d = 0.687,GO= 51.4/cm, Ps= 41.2 roW,andal ---49.6/cm. (The value of I0 is arbitrary
but we used I0 = 10 mA for the computed GO above. Note that GO and I0 can be combined
to form a single constant.) The resulting gain curves for these parameters are also
illustrated in Fig. 2. It is interesting to note that the value of c is not unity due to the fact
that bimolecular recombination is significant. The value of d is unity for a classical two-
level system which is broadened homogeneously, whereas, it is 1/2 for an
inhomogeneously broadened laser. Ps is the saturation power and al is the modal loss
coefficient in the absence of gain. At threshold, the intracavity power P = 0, and the
threshold current is
(Gth + al) TMIth=I 0_, GO (7)
where Gth = (1/2L) ln(1/R1R2), and R1 and R2 are the facet reflectivities. In the event the
reflectivities are high, the intracavity power is almost constant along z and the gain G =
Gth. The emission power from the R1 facet is is'obtained from (6)
rl-R,,TV(.I,Tc/dI-1Prad=Ps 1-V- lJ[."_] (8)
63
This is just the expression for the emission power versus drive one would obtain for
uniform photon densities. When one or both of the facet reflectivities is small, the
intracavity power varies considerably along the laser axis and must be computed from the
integral equation
1 Z(9)
where P0 is an eigenvalue, facet 2 lies at z = 0 and facet 1 lies at z = L. The boundary
condition on G(P) requires its integral over the length of the cavity be 1/2 ha(1/R1R2).
Figure 3 shows the results for a laser with L = 250 I.tm and an emission power, Pracl
= 50 roW. The back facet has a reflectivity R2 = 1, and the output facet reflectivity R1 is
treated as a parameter. (The threshold current can be computed using Eq. (7).) When the
output facet reflectivity is 0.3, P is almost constant along the laser axis; however, for R1 =
0.05, the intracavity power varies fi'om about 24 mW at z = 0 to about 55 mW at z = L.
Figure 4 shows power versus current for the different reflectivities.
In conclusion we have developed the self-consistent model for semiconductor lasers
(using the CSP-DH laser structure as an example) without assuming constant optical power
along the laser axis. The advantage of the present approach allows for the analysis of
lasers (typically high-power) with small facet reflectivities which produce nonuniform
photon densities along the propagation direction. Generally, the hole burning effects will
be larger at the output facet because the optical density is highest there. Further, hole
burning is nonuniform in the direction of propagation. However, in long contemporary
lasers these nonuniformities are slow compared to the carrier diffusion length. We also
have developed analytical equations for the modal gain coefficient, the threshold current
density, and (in the limit of high facet reflectivities) the radiated power for a specific CSP
laser structure.
64
The authors wish to thank R. Amantea, M. Ettenberg and M. Lurie for many
helpful technical discussions.
.
.
3.
.
.
.
7.
8.
.
10.
11.
12.
13.
14.
15.
16.
65
REFERENCES
G. P. Agrawal, W. B. Joyce, R. W. Dixon, and M. Lax, Appl. Phys. Lett., 43,
11, 1983.
tt Baets and P. E. Lagasse, Electron. Lett. 20, 41, 1984.
P. Meissner, E. Patzak, and D. Yevick, IEEE J. Quantum Electron. QE-20, 899,
1984.
R. Baets, J. P. van de Capelle, and P. E. Lagasse, IEEE J. Quantum Electron. QE-
21,693, 1985.
J. K. Butler, G. A. Evans, B. Goldstein, IEEE J. Quantum Electron. (to be
published in November 1987).
J. Buus, IEEE J. Quantum Electron., QE-15, 734, 1979.
W. Streifer, D. R. Scifres and R. D. Burnham, Appl. Phys. Lett. 37, 877, 1980.
W. Streifer, D. R. Scifres and R. D. Burnham, IEEE J. Quantum Electron. QE-17,
1521, 1981.
W. Streifer, D. R. Scifres and tt D. Burnham, IEEE J. Quantum Electron. QE-17,
736, 1981.
M. Ueno, IL Lang, S. Matsumoto, H. Kawano, T. Furuse, and L Sakuma, IEE
Proc., 129, Pt. I, 218, 1982.
S. Wang, C. Y. Chen, A. S. Liao, and L. Figueroa, IEEE J. Quantum Electron.
QE-17, 453, 1981.
IC A. Shore, Opt. and Quantum Electron. 15,371, 1983.
G. P. Agrawal, IEEE J. Lightwave Technol. LT-2, 537, 1984.
J. Buus, IEE Proc. 132, 42, 1985.
R. Papannareddy, W. E. Ferguson, and J. tC Butler, IEEE J. Quantum Electron
(Submitted for publication).
A. Yariv, Quantum Electronics, 2nd Edition, John Wiley &Sons, Inc., New York,
1975.
66
FIGURES
Fig. 1
"Fig. 2
Fig. 3
Fig. 4
TheCSP-DHgeometryusedfor the laserdevice.
Opticalgaincharacteristicsof thefundamentalmodeasafunctionof theintracavitypower. Thedrive currentis for adeviceof lengthL = 100Ixm.The dotsare
computedfrom aself-consistentmodel5andthesolidcurvesareobtainedfrom
Eq. (6).
Theintracavitypowercomputedfrom Eq.(9) for a laserof lengthL = 250pan.
ThebackfacethasR2= 1while thefront facetreflectivity is treatedasa parameter.
The totalemissionpowerfrom thefrontfacetis 50 roW.
Emissionpowerfrom thefront facetversusdrive current Thisdatawasobtained
from repeatedsolutionsof Eq. (9).
67
II II
_..J
!
ILl
..,--I
I-X-I
m
Im
I=
r,_
68
I.JI./I. ]lj
g
//!IZ :.
///ii" //i
./././,. °(T-mo) O 'IN]I:)IJ:I]O:) NlVO1V(lOfl
69
I I
a,,i4
if
(Mm) d '_l]MOd A.LIAV_)VII.LH!
0
c",4
mc,,, 4
c",,,f
m,a,,,i
Om
Q
,I,,,4
I i.,.C-j
I.,r2
O
OC",,I
s=
P,,,,,I
ii
i,.,..m
__1
mmim
TO
o
!....I
I..-I
00
0
E
I.--p..
_G
ZD
mmm
¢DI.n
¢D
II
I I I t0 0 0 0
(M'") H]MOd NOISSIfl]
Appendix B
73
Observations and Consequences of Non-UniformAluminum Concentrations in the Channel Regions
of AIGaAs channeled-Substrate-Planar Lasers"
Gary A. Evans, Bernard Goldstein" and Jerome K. Butler""David Sarnoff Research Center
CN 5300
Princeton, NJ 08543-5300
ABSTRACT
Compositional changes in the n-clad layer within the channel region of channel
substrate planar (CSP) type semiconductor lasers have been observed. As a consequence,
a large optical cavity (LOC) or an enhanced substrate loss (ESL) version of the CSP
geometry may result, both of which may have significantly different characteristics from
those of a conventional CSP laser. The CSP-LOC generally has a larger near field spot
size, while the ESL-CSP is characterized by an off-axis, asymmetric far-field pattern.
*This work was supported in part by the National Aeronautics and Space Administrationunder Contract Number NAS 1-17441.
**Presently with Solarex Corp., Newtown, PA 18940
***Southern Methodist Univ., Dallas, 'IX 75275
PRI_CI_INO PAG_ BLANK NOT I_ILMED
75
Observations and Consequences of Non-UniformAluminum Concentrations in the Channel Regions
of AIGaAs Channeled-Substrate-Planar Lasers"
Gary A. Evans, Bernard Goldstein** and Jerome K. Butler***David Sarnoff Research Center
CN 5300Princeton, NJ 08543-5300
I. INTRODUCTION
Generally, in the design and analysis of liquid-phase-epitaxy grown channeled-
substmte-planar (CSP) lasers (1,2), the A1 concentration within each epilayer is assumed to
be uniform. Experimentally this is often not the case: within the channel of the n-clad
layer (Fig. la), significant non-uniformities in the direction perpendicular to the junction
can exist in the A1/Ga ratio. This paper presents and discusses this evidence, and
examines the consequences of these non-uniformities. We find that the AlAs non-
uniformities in the channel can change a conventional CSP double heterostructure (DH)
into either a CSP-LOC (large optical cavity) (3,4) or a CSP structure with increased losses
in the substrate which we have called an Enhanced Substrate Loss (ESL) CSP. The
resulting CSP-LOC laser generally has a wider perpendicular full-width-half-power
(FWHP) near-field distribution, and similar or larger perpendicular far-field beam
divergence compared to a conventional CSP laser. The ESL-CSP laser has an asymmetric
perpendicular far field and can have either a larger or smaller FWHP perpendicular far field
compared to a conventional CSP laser.
*This work was supported in part by the National Aeronautics and Space Adminstrationunder contract number NAS 1-17441.
"*Presendy with Solarex Corp., Newtown, PA 18940°
***Southern Methodist Univ., Dallas, TX 75275
PRECI_)ING PAGE BLANK NOT FILMED
76
Asymmetries in the perpendicular far fields of conventional DH lasers are
theoretically unexpected: The largereal index steps perpendicularto the junction in
AIGaAsDH lasers(Fig. la) requirethatthenear-field solution to the electromagnetic wave
equation be almost real with only a negligible imaginary component due to active layer gain
and material losses. Since the far field pattern is related to the near field distribution by a
Fourier transform, the far field pattern should be symmetric about an axis normal to the
laser facet because the Fourier transform of a real function is symmetric. However,
double-heterostructure lasers with a thin (< 1.0 gin) cladding layer will have asymmetric,
off-axis far-fields (5,6,7) due to radiation losses in the cap or substrate.
Experimental measurements of our CSP lasers indicate that in some channels the
gradient is from high A1 to low A1 starting at the bottom of the channel, while in other
channels on the same wafer, the gradient is reversed. Both types of aluminum composition
grading can occur in the channels of CSP lasers from portions of a wafer that otherwise
produce conventional CSP-DH lasers. Aluminum composition grading has been observed
in channels etched in both the V-groove (Fig. 2a) and the dovetail (Fig. 2b) directions.
II. EXPERIMENTAL MEASUREMENTS AND ANALYSIS
A. Higher AlAs Concentration at the Bottom of the Channel
The principal experimental technique used for the surface compositional analysis on
the cleaved facets of the lasers was Auger Electron Spectroscopy using a primary electron
beam with a resolution of about 1000/_. Figure lb shows Auger spectra that indicate the
surface composition on a cleaved facet of a CSP type laser at the points x = 0.3 _m and x =
1.2 tzm shown in Fig. la. The magnitudes of the Ga, As and A1 lines shown on the
spectra reflect the concentration of these constituents at the two points and indicate that the
AI concentration at the bottom of the channel is about twice that just below the active
region. Note that the change in the magnitude of the A1 line is tracked by a corresponding
change in the magnitude of the Ga line while the As line has remained essentially
77
unchanged. Examining a random sampling of CSP lasers has establishedthat the
magnitudeof theconcentrationvariationsindicatedin Fig. lb rangesfrom zeroto abouta
factor of two.
The A1concentrationvariationwearereportingcanalsobeseenqualitatively in the
scanningelectronmicrographof thecleavedfacetshownfor adovetailchannelin Fig. 2b.
Here,partof thebackscatteredelectronsignalisdueto theaverageatomicnumberZ of the
surfaceunder examination. Thus, brighter regions in the micrograph are re_ons of
material with a higher atomicnumbercomparedto neighboringdarkerregions,so that
re_onsof higherA1concentrationwill bedarkerthanthoseof lowerA1 concentration. The
difference in atomic number between AI and Ga is 2.5. In the SEM micrograph of Fig. 2b,
the laser structure is clearly defined. The 0.9 p.m cap layer and 0.1 p.m active layer are
delineated by the bright horizontal images (0 and 7% AIAs content, respectively), while the
two cladding layers (both nominally 33% ALAs) show up as darker regions on either side
of the active layer. Within the channel shown in Fig. 2b, the dark region indicates a
significantly higher AlAs content at the channel bottom, while the lighter region indicates a
significantly lower AIAs content adjacent to the active layer--in agreement with the Auger
analysis data shown in Fig. lb.
A non-uniform A1/Ga ratio within the channel will effect the dielectric profile
perpendicular to the junction. Figure 3 contains a series of possible index profiles together
with their electric field distributions for a) a uniform AI/Ga ratio in the channel (CSP); b) a
higher A1/Ga ratio at the channel bottom (CSP-LOC); and c) a lower AI/Ga ratio at the
channel bottom (ESL-CSP). Graded index profiles are also possible. However, their
characteristics are qualitatively similar to the abrupt step profiles discussed in this paper.
The aluminum concentrations and corresponding index values for the three structures
shown in Fig. 3 are tabulated in Tables 1, 2, and 3. The loss of each epi-layer is taken to
be 10 cm -1. The value used for the substrate loss (xs (at k = 0.83 _tm), for all calculations
78
was5000cm-1 (8). However, thereis insignificantdifferencein all calculatedresultsas
c_svariesfrom 1000to 10000cm"1, in agreementwith earlierreports(2,9).
The electricfield distributionperpendicularto thep-njunction in thechannel,Ey(X),
is obtainedby solvingtheonedimensionalwaveequation(assumingregionA of Fig. la
is infinitely wide):
_2Ey/0X2 + (erko2 - 132)Ey= 0 [1]
with the usual boundaryconditions (1,2) using an algorithm for calculating complex
modesin plane-layered,complexdielectric structures(10). Here,er(X) is thecomplex
relativeelectric permittivity, ko = 2rr/'Lo, andXo is the free space wavelength. An exp[-
i([3z-c0t)] longitudinal and time variation of the electric field is assumed. The real part of the
complex index of refraction n*(x), shown in Fig. 3, is the real part of the square root of the
complex relative electric permittivity. The effective index of region A (Fig.la) is
nAeff = Re([3/ko) [2]
and the effective absorption coefficient of region A is
=-2ko Im(13/ko) [31
The theoretical calculations of the electric field and near field intensity distributions
are both normalized according to
OQ
1 = I Ey(x) E*y(X) dx [4]
79
The far field intensitydistribution I(0) is calculatedfrom the complexelectricfield
distributionaccordingto
,, d 12I(0) = I g(0) IEy(x) exp(j sin0 kox)cLx]"/lg(0) _Ey(x) [5]-OO -OO
where g(0) is the obliquity factor (11).
The confinement factor F is def'med for any layer as
oo
rlayer = IEy(x) E*y(x)dx/IEy(x) E*y(X) dx [61layer -¢_
The data in Figs. 1 and 2 indicate cases in which the aluminum concentration is
highest in the bottom of the channels. In these cases, the resulting perpendicular index
profile is no longer that of a simple double heterostructure (Fig. 3a), but that of a large
optical cavity (Fig. 3b). Such a dielectric profile was originally introduced intentionally
(3) to lower the optical power density by producing a wider perpendicular near-field
distribution than that of a conventional DH configuration. In Fig. 4a the FWHP of the
near field intensity is plotted as a function of AlAs composition (or index at _. = 0.83 gm)
of the LOC layer assuming the AlAs mole fraction of the p- and n-clad layer is 0.33 and
that of the active layer is 0.07. The CSP-LOC FWHP near field intensity of the
fundamental mode reaches a maximum of about 0.4 gm (for a 0.4 gtm thick LOC layer) and
0.72 gtm (for a 0.9 gm thick LOC layer) at AlAs compositions of the LOC layer of about
18%. These curves are not meant to be design curves for LOC structures. They are meant
to indicate general patterns of behavior and do not consider, for example, higher order
modes which are possible in very thick (> 0.5 gin) LOC layers (3,4).
80
Parenthetically,if thealuminumcontentis greateratthebottomof thechannelthanin
the _ layer, it is possiblefor a LOC layer of a CSP-LOC to actually decrease the
FWHP of the near field intensity. This is shown in Fig. 5a, which is a plot similar to that
in Fig. 4a except that the n-clad region has an AlAs mole fraction of 45%: the FWHP of
the near field intensity initially decreases as the AI is increased in the LOC layer. In both
Figs. 4 and 5, as the AlAs composition of the LOC layer is decreased below 33%, the
CSP-LOC FWHP near field spot size is always greater than that for a conventional CSP
laser.
Figure 6 contains experimentally observed photographs of the near-fields of three
CSP lasers operating at 20 mW cw, all from the same wafer, showing considerable
variation in the perpendicular spot size. The near fields in Fig. 6b and c show some
asymmetry in the direction of the channel, which is not characteristic of a conventional CSP
DH with equal AlAs compositions in the p- and n- clad layers. These experimental results
are in qualitative agreement with the calculated near-fields shown in Fig. 7a for a
conventional CSP laser, a CSP-LOC laser (LOC composition =22% AIAs, n-clad
composition = 33% AlAs), and an ESL-CSP laser (ESL-LOC layer composition =33%
AIAs, ESL n-clad layer composition = 26%). The far fields corresponding to the near
field intensity and phase distributions of Fig. 7a and b are shown in Fig. 7c.
An additional consequence of aluminum composition changes in the channel region is
a variation in the FWHP perpendicular beam divergence. The FWHP of the far-field
radiation lobe of the laser corresponding to Fig. 6(a) is 350, while that of the laser
corresponding to Fig. 6(c) is 28 °. While this range in FWHP perpendicular beam
divergence could be explained by a variation in active layer thickness from 650 ,_, to 1000
]k for a conventional CSP laser, an alternative explanation is a nonuniform AI compositon
in the channel region as shown in Figs, 5b and 8. In Fig. 8 the FWHP perpendicular beam
divergences are plotted (for the same parameters used in Fig. 4 to plot the FWHP of the
near field) as the A1 composition of the LOC layer (CSP-LOC structures) or n-clad layer
81
(ESL-CSPstructures)variesfrom 33% to7%. Figure8 showsthatnonuniformiriesin the
AI composition in the channelregion cancausea rangein FWHP perpendicularbeam
divergencesfrom about20° to morethan50o for afixed active layer thicknessof 800 ,_.
These calculations indicate that observed variations in FWHP perpendicular beam
divergences can result not only from variations in active layer thicknesses, but also from
variations in A1 compositions in the LOC or n-clad layers of any semiconductor laser.
Finally, the larger perpendicular near-field spot sizes shown in Fig. 6 (and calculated
in Figs. 4 and 5) result in a decreased optical power density which might be expected to
result in longer operating life at moderate to high power. This expectation was found
experimentally for the three lasers (all from the same wafer) whose near-field photos axe
shown in Fig. 6. Operating at 20 mW cw and at 30 ° C, the flu:st unit (6a) lasted 50
hours, the second unit (6b) about one thousand hours, while the third (6c) is still operating
after 17,000 hours.
B. Lower AlAs Concentration at the Bottom of the Channel
Shown in Fig. 9 is a schematic drawing of the channel re,on of a CSP laser together
with A1 concentrations as given by Auger Electron Spectroscopy data. Here, the AI content
is lower at the bottom of the channel. The resulting index profile (Fig. 3c) can be thought
of as "pulling" some of the mode power into the substrate, increasing the mode loss. This
mechanism, which we call Enhanced Substrate Loss (ESL), can also be explained by
realizing that all of the modes supported by the index profiles shown in Fig. 3 are complex
modes (12) (i.e., the longitudinal and transverse wave vectors have a real and imaginary
component), because the field solutions to the electromagnetic wave equation [1] axe
"sinusoidal" in the substrate (13). Usually, the n-clad region separating the active region
(and LOC layer, if present) from the substrate is thick enough (> 1 I.tm) that field
penetration of the laser mode into the substrate is negligible (see Fig. 3a, b). However, if
the mole fraction of AlAs is lower at the bottom of the channel than at the top of the
82
channel,the higher index portion (layer 4 of Fig. 3c) in the channelacts like an anti-
reflection coating(14)betweenthehighaluminum,low indexportion(layer 3) andthe no
aluminum,very high index substrate(layer5), therebycouplingor redistributing alarger
fraction of themodepower into the substrate.In Fig.10, theratio of energyconfined in
the lasersubstrateto thetotalmodeenergy(I-"s, thesubstrateconfinementfactor) is plotted
asa function of the mole fractionof AlAs (or indexof refractionat _ = 0.83 I-tin)of the
bottom half of the channelregion (layer4) for the indexprofile shownin Fig. 3c. The
peaksin this plot correspondto indexvaluesthat optimizethecoupling of light into the
substmtefor the0.9 grn thicknessof the index"matching"layer(layer4). In thevicinity
of thesepeaks,themode has significant energy directed both along the waveguide axis and
into the substrate, and the Corresponding far field pattern is no longer peaked perpendicular
to the facet, but is tilted towards the substrate on the order of a few degrees as shown by
the calculated far fields--see Fig. 7c and insets in Fig. 10. This far-field tilt is consistent
with the asymmetric near-field phase of the ESL-CSP laser (Fig. 7b) which, unlike that of
the CSP-LOC or CSP laser, is not flat over the region (0.4 < x < 2.0) where the optical
field has a significant amplitude. The asymmetric far field pattern for the ESL-CSP laser in
Fig. 7c is caused by the asymmetry of the near field phase variation in Fig. 7b.
For both the ESL-CSP and CSP-LOC lasers, the thicknesses and refractive indices of
the LOC and n-clad layers chosen for the theoretical models are not unique. The cases
illustrated in this paper correspond to fixed layer thicknesses with a varying index of
refraction in one layer (Fig. 4, 5, and 10). However, a curve similar to Fig. 10 is obtained
by assuming an index n4 (n3 <n4 <n5) for the n-clad layer and varying the thickness of the
n-clad layer.
Experimentally observed far-field intensity patterns of several ESL-CSP lasers (with
V-groove channels) from the same wafer are shown in Fig. 11. The experimental
asymmetric profiles agree qualitatively with the theoretical profiles shown in Fig. 10. In
general, the asymmetric far field profile for a given ESL-CSP laser does not change as a
83
function of power even at power levels above 100mW. The shift of the peak of the
experimentalfar fields from thenormalto thefacetrangesfrom lessthan2° to about9° , in
agreementwith thecalculatedshifts(up to 8°) shownin Figs.7cand10. Theexperimental
rangein FWHP of thefar field patternsrangefrom 18° to 33° comparedto a theoretical
rangeof 20° to 42° for theparametersassumedin Fig.8. Distinct sidelobesappearin the
experimentalfar field shownin Fig. 1la andin someof thecalculatedfar fields shownin
Fig. 10and7c. Thesesidelobescanbeexplainedastheinterferencepatternin thefar field
betweentwo peaksof thenearfield intensity. For example,thenearfield distribution of
theESL-CSPlasershownin Fig. 7acanbeapproximatedby two point sourcesseparated
by 1.3gin. From Fig. 7b, thesmallerpeakis shiftedin phaseby anaverageof about90°
relative to the main peak. The interference pattern from this simple two point
approximationhasa central peak at 9.2° andsidelobesat 28.6° and -52.9°, in general
agreementwith theESL-CSPfar field patternshownin Fig.7c.
Asymmetriesandradiation patternshifts in theperpendicularfar field have been
observedin many types of AlGa.As semiconductorlasers and we suggestthat non-
uniform A1compositionsareoftentheexplanation.
III. DISCUSSION
The cause of a non-uniform A1/Ga ratio occurring within the channel of a CSP laser
during the LPE growth of the f'Lrst cladding layer is not fully understood. However, we do
know 1) that the layer growth must be faster at the bottom of the channel than at the top in
order to result in flat growth prof'fles, 2) that there must be a lateral component of growth
within the channel as well as a perpendicular component, 3) that the wall and bottom
curvatures of the channel present crystallographic planes to the growth nucleation that are
different from those above the channel where the growth is planar (15), and 4) that there
can be varying degrees of meltback at the channel walls and shoulders as shown in Fig. 2.
Any of these conditions can readily affect the composition of the ternary compound that
84
initially nucleates and freezes out from the A1GaAs melt, and can alter this composition as
the growth proceeds and overall growth conditions change. Furthermore, if the AI/Ga
ratio changes during the inidal growth, then changes in the local melt composition may
occur which might further change the A1/Ga ratio later along in the growth. From the
A1As-GaAs phase diagram (16), very small changes in A1 in the melt produce very
significant changes in the AI content in the grown material. The basis for a non-uniform
A1/Ga ratio within the channel of a CSP laser discussed above is also consistent with the
variability of this effect from channel to channel since it would be the local conditions
around each channel (shape, initial freeze-out rate, local melt composition) that would
determine the magnitude of the effect. The lasers discussed in this paper were grown at
800°C with about 4°C of supersaturation and a cooling rate of 0.75°C_./minute. The p- and
n- dopants were Ge and Sn.
While the CSP-LOC structure produced by higher A1 concentrations at the bottom of
the channels has generally familiar and well understood consequences, the ESL-CSP
sn-ucture produced by lower A1 concentrations at the bottom of the channel has some
peculiar properties. First, the internal losses of the perpendicular component of the ESL
mode can be almost an order of magnitude higher than that of the conventional CSP or
CSP-LOC_ as shown in Fig. 12, which could noticeably increase the threshold current and
decrease both efficiency and lifetime. (However, we have not seen a clear correlation
between increases in threshold current and decreases in differential quantum efficiency with
asymmetries in the far field pattern of our lasers.) Second, the near field and far field
FWHP plots summarized in Figs. 4,5,7 and 8 are not those intuitively expected from plane
wave, uniform intensity diffraction theory which equates the far field beam divergence 0 B
to L/D, where D is the near field aperture: The ESL-CSP structure can have a narrower
FWHP perpendicular near field intensity and a narrower FWHP perpendicular beam
divergence than a CSP or CSP-LOC structure. Additionally, while the FWHP of the
• perpendicular near field spot sizes for the CSP-LOC structures are 2 to 3 times larger than
85
that for a conventional CSP (Fig. 4), the CSP-LOC FWt-IP beam divergences are larger by
typically 10 ° (Fig. 8). Finally, both the CSP-LOC FWHP beam divei'gences (Fig. 5b and
8) and the FWHP near field intensifies (Fig. 4a and 5a) simultaneously increase over some
ranges of ALAs variation. These apparent contradictions result because the FWHP, by
itself, of the various near field intensities can be a misleading and inappropriate
representation of the near field aperture. For example, the ESL-CSP structure has a very
narrow FWI-IP beam divergence of about 24 ° (Fig. 7c) and a narrow FWFIP near field
spot size of about 0.26 _.m. However, because of an irregular distribution of mode power,
the effective near field aperture (Fig. 7a) is actually about 1.5 _tm. The reason for these
apparent contradictions is primarily because the FWI-IP of the near field intensities
contain varying fractions of the mode power. Such apparent contradictions would be
largely eliminated (17) by a definition of near field aperture and far field beam divergence
corresponding to near and far field widths encIosing the same percentage (for example,
80%) of mode power, instead of the experimentally convenient measurement of FWHP.
The effect of A1 variations on lateral mode confinement has not been investigated in
detail. However, since only the fundamental lateral mode is observed experimentally, the
difference between the magnitude of the complex effective index in the channel (region A
of Fig. la) and outside the channel (region B of Fig.la), Aneff (= [n*Aefft - In*t3effl),
must be much less than 0.1 for channel widths ranging from 4 to 7 t.tm. Figure 13 shows
that while the effective index in the channel region (n*Aeff) for the ESL-CSP structure
changes negligibly with changes in A1 concentration (- 4 x 10-4), the effective index in the
channel reron of the CSP-LOC smacture changes significantly (- 2 - 6 x 10-2). Thus, the
observed absence of higher order lateral modes, for the CSP-LOC structure, suggests that
A1 composition changes may occur in the n-clad layer _ the channel that follow, at
least partially, those inside the channel, and result in a corresponding increase in n*Beff.
86
IV. CONCLUSIONS
The net result of the effect described here is that due to compositional changes in the
n-clad layer within the channel, a CSP-LOC or ESL-CSP can be inadvertently grown
instead of a conventional CSP laser, all three of which have significantly different
characteristics. We also give a theoretical explanation for asymmetric perpendicular far-
field patterns which have been experimentally observed in CSP and other AlGa.As laser
structures.
ACKNOWLEDGEMENTS
We are indebted to D. Szostak for the Auger surface analysis measurements, N.
Dinkel for the growth of the lasers, V. Masin for computer program development and
technical discussions, E. DePiano, D. Tarangioli, and M. Harvey for laser processing, and
to D. Cartin, N. Carlson, I. Connolly, M. Ettenberg, J. Hammer, F. Hawrylo, M. Lurie,
and S. Palfrey for technical discussions.
87
REFERENCES
1) K. Aiki, M. Nakamura, Takao Kuroda, J. Umeda, R. Ito, Naoki Chinone, and M.
Maeda, "Transverse Mode S tabilized AlxGal_xAs Injection Lasers with Channeled-
Substrate-Planar Structure," IEEE Journal of Quantum Electronics, Vol. QE-14, No.
2, pp.89-94, Feb. 1978.
2) T. Kuroda, M. Nakamura, K. Aiki, and J. Umeda, "Channeled-substrate-planar
structure AlxGal_xAs lasers: an analytical wavegq.fide study," Applied Optics, Vol.
17, No. 20, pp. 3264-3267, 15 October 1978.
3) H. Kressel and J. K. Buffer, Semiconductor Lasers and Heterojuncfion LEDs,
Chapter 7, Section 5, Academic Press, New York, pp. 230-234, 1977.
4) Toshiro Hayakawa, Takahiro Suyama, Hiroshi Hayashi, Saburo Yamamoto, Seiki
Yano, and Toshild Hijikam, "Mode Characteristics of Large-Optical-Cavity V-
Channeled Substrate Inner Stripe Injection Lasers," IEEE Journal of Quantum
Electronics, Vol. QE-19, No. 10, pp. 1530-1536, October 1983.
5) I. K. Buffer, H. Kresset, and I. Ladany, "Internal Optical Losses in Very Thin CW
Heterojunction Laser Diodes," IEEE Journal of Quantum Electronics, Vol. QE- 11,
No. 7, pp. 402,408.
6) William Streifer, Robert D. Burnham, and Donald R. Scifres, "Substmte Radiation
•Losses in Ga.As Heterostructure Lasers," IEEE Journal of Quantum Electronics, Vok
QE-12, No. 3, pp. 177-182, March 1976.
88
7) D. R. Scifres, W. Streifer, and R. D. Burnham, "Leaky wave room-temperature
double heterostructure GaAs:Ga.AIAs diode laser," Appl. Phys. Lett., Vol. 29, No.
1, pp.23-25, July i976.
8) H. C. Casey, Jr., D. D. Sell, and K. W. Wecht, "Concentration dependence of the
absorption coefficient for n- and p-type GaAs between 1.3 and 1.6 eV," J. AppI.
Phys., vol. 24, pp.63-65, Jan. 1974.
9) D. R. Scifres, R. D. Burnham, and W. Streiffer, "Output coupling and distributed
feedback utilizing substrate corregations in double-heterostrucmre Ga.As lasers,"
Appl. Phys, Lett., Vol. 27, No. 5, pp.295-297, 1975.
10) R. B. Smith and G. L. Mitchel, "Calculation of Complex Propagating Modes in
Arbitrary, Plane-Layered, Complex Dielectric Structures. I. Analytic Formulation.
II. Fortran Program MODEIG," EE Technical Report No. 206, University of
Washington, Seattle, Washington.
11) L. Lewin, "Obliquity-factor correction to solid-state radiation patterns," Journal Of
Applied Physics, Vol. 46, No. 5, pp. 2323-2324, May 1975.
12) Theodor Tamir and Foon Yeon Kou, "Varieties of Leaky Waves and Their Excitation
Along Multilayered Structures," IEEE Journal of Quantum Electronics, Vol. QE-22,
No. 4, pp. 544-551, April 1986.
13) This feature of the perpendicular waveguide mode for x > 0.38 outside the channel
(region B in Figure la) is the reason CSP lasers have a higher effective index inside
the channel region than outside.
89
14) M. Born andE.Wolf, Principles of Optics, Chapter 1, pp. 61-66, Pergamon Press,
New York, 1975.
15) J. W. Calm and D. W. Hoffman, "A Vector Thermodynamics For Anisotropic
Surfaces--II. Curved and Faceted Surfaces", Acta Metallurgica, Vol. 22, pp. 1205-
1214, October 1974.
16) H. Kressel and J. K. Butler, Semiconductor Lasers and Heterojunction LEDs,
Chapter 11, p. 372, Academic Press, New York, 1977.
17) The far field beam divergence from a fixed aperture D will, however, have slight to
moderate variations as either the near field intensity distribution or the near field phase
is changed.
9O
FIGURE CAPTIONS
Figure 1. a) Geometry of a typical CSP type laser; x = 0 is the top of the active layer
and x = 1.8 gm is the bottom of the channel, b) Auger analysis of a cleaved
facet of a CSP type laser showing a higher aluminum composition near the
bottom of the channel (x = 1.4 lain, dashed line) than near the top of the
channel (x = 0.4 Izm, solid line).
Figure 2. Scanning electron micrographs of a cross section of a CSP type laser with the
channel etched in a) the V-groove direction, and b) the dovetail direction. The
dashed black Lines indicates the channel profiles before growth.
Figure 3. Index profiles ( .... ) and corresponding electric field distributions (--) for a)
a conventional CSP laser, b) a CSP-LOC laser; and c) an ESL-CSP laser.
The layer compositions, thicknesses, and effective index for each structure are
listed in Tables 1-3. The dashed rectangles in a) and b) show the field
distributions on expanded scales for x > 1.8 gin.
Figure 4. a) Calculated near field FWt-IP as a function of the % AIAs (or index of
refraction at _. = 0.83/am) of a 0.4 lain ( .... ) and 0.9 ]am (--) thick LOC
layer (CSP-LOC geometry) or of the 0.9 [am ( .... ) thick n-clad layer (ESL-
CSP geometry). The n-clad layer (layer 4, Figure 3b) for the CSP-LOC layer
has an AIAs mole fraction of 0.33, and the LOC layer (layer 3, Figure 3c) for
the ESL-CSP has an A1 As mole fraction of 0.33. b) The calculated near
field FWHP as a function of AlAs or index of refraction for the ESL-CSP
structure on an expanded scale. The common point to all three curves (at an
91
indexvalue= 3.40657)correspondsto theconventionalCSPlaserdescribed
in Table 1.
Figure 5. Calculateda)nearfield FWHP andb) far field FWHP asafunction of the%
AlAs (or indexof refractionatk = 0.83 I.tm)of a 0.4 _tm(.... ) anda0.9 I.u'n
(--) thick LOC layerof a CSP-LOCwith ann-clad AIAs mole fraction of
0.45.
Figure 6. Near field microgaphs of CSP type lasers operating at the same power
output showing si_ificantly different perpendicular near fields.
Figure 7. Calculated a) near field intensities, b) near field phases, and c) far field
intensifies perpendicular to the junction for a conventional CSP laser (----), a
CSP-LOC laser ( .... ), and an ESL-CSP laser ( .... ) for the parameters listed in
Tables 1-3.
Figure 8. Calculated far field FWHP as a function of the % AlAs (or index of refraction
at ),. = 0.83 gin) of a 0.4 p.m (---) and 0.9 ktm (--) thick LOC layer (CSP-
LOC geometry) or of the 0.9 la.m ( .... ) thick n-clad layer (ESL-CSP
geometry). As in Figure 4, the n-clad layer for the CSP-LOC layer has an
AIAs mole fraction of 0.33, and the LOC layer (layer 3) for the ESL-CSP has
an A1 As mole fraction of 0.33. The common point to all three curves (at an
index value = 3.40657) corresponds to the conventional CSP laser described
in Table 1.
92
Figure 9. Composition measured by Auger analysis at four positions along a cleaved
facet of a CSP type laser showing a lower aluminum composition near the
bottom of the channel than near the top of the channel.
Figure 10. The substmte confinement factor F s as a function of the % AiAs (or index of
refraction at _, = 0.83 btm) of the n-clad layer at the bottom of the channel of
an ESL-CSP laser. The inset far field intensity vs angle patterns show a
large variation in asymmetry as a function of % AlAs of the n-clad layer.
Figure 11. Experimentally measured far field intensity patterns for ESL-CSP lasers.
Figure 12. The perpendicular mode loss in region A of Figure la as a function of the %
AlAs (or index of refraction at _, = 0.83 I,n'n) of the n-clad layer at the bottom
of the channel of an ESL-CSP laser. The right hand ordinate is the imaginary
part of the' effective index.
Figure 13. The calculated effective index in the channel as a function of the % AlAs (or
index of refraction at _. = 0.83 g.m) of a 0.4 I,tm ( .... ) thick LOC layer
(CSP-LOC geometry) or of a 0.9 _m ( .... ) thick n-clad layer (ESL-CSP
geometry). The n-clad layer for the CSP-LOC layer has an AlAs mole
fraction of 0.33, and the LOC layer (layer 3) for the ESL-CSP has an AI As
mole fraction of 0.33. The common point to both curves (at an index value =
3.40657) corresponds to the conventional CSP laser described in Table 1.
The effective index of the ESL-CSP laser changes by less than 4 x 10 -4 over
the index range of the n-clad layer shown. Also shown are typical design
values for the effective index in the channel nA(CSP)ef f and outside the
channel nB(CSP)eff for a conventional CSP laser.
g3
_ x
_ E
q_ oo _ L_
_D
c_
r_
i-u
c._ oo
_ A
g4
E
_b
X
,,6
[.-.,r,D
r,DO
I
rj_
N
_1
F,
_b
E--1
c_
X
o,-i
E:::I.
°_
i._.
A I
g5
E
rj _
c_I..
2_
_ E_ m
E- x
.,.-q
EzL
20
_0w
Z0
(,9
Z0
(Z:
g6
(SIINrl '8_1_'} 7_'N01$ W30N'9
E E::L ::L
Od -
u= _11
0 m
...I
E
E:k
I E! ::k
_ 04
97
ORIGINAL PAGE ;S
OF POOR QUALITY
98
ORIGINALPAGE ISOF POORQUALITY
99
NOI.LZ)V_..-13__40 X30NI
r-7
-°D
_ _ X
-o
-0-o
--c_I
0III I --
I
dI
0d
10731__ Z)I_I.LZ)373] a_
100
NOllOV_A3_ AO X3(]NT
07314 91NJ.:3373] 08
101
NOI/3V8338 30 X3qNI
(1731=I 31_119373] aH
>-I,-
0.8ZIJ.l
t- 0.7ZI-t
0.6_JbJ
ILl
z 0.4b.l-t-l--
,, 0.30(3.-r- 0.2
I.i.
30 25
102
% ALAs
20 15 I0
I
•--. 0.9H.m LOC
O.4/._.m LOC..... ESL-CSP
CSP
I I I
_rLl_SP- LOC
:3.40 3.45 5.50 3.55 5.60
INDEX OF LOC LAYER (CSP-LOC) OR
ESL n-CLAD LAYER (ESL-CSP)
0
>-!--03Zb,.II--ZH
C3._JI.IJm A
" E
LIJZ
LI.I"t-I--
h0
O.
I.I.
0.25625
0.25600
0.25575
0.25550
0.25525
m
i
i
m
m
i
m
m
i
_o q)
i
% ALAs
50 25 20 15 10I I I I I
%e, t
iP o
I"Ib
; •
: : _.rl__sL= • -CSP
. . /•I • °°°eo
• • •
_ oo •
• • eo e°°e° et'Dooo
_ °°°°eo°° o
o;oqb
0.255003.40
0
l __
g
u
m
i
-,,,, I,,, ,I,, ,,I,, II ,, ,-5.45 3.50 3.55 5.60 3.65
INDEX OF ESL n-CLAD LAYER (ES'L-CSP) 6540
103
_- 0.8
0.7
0.6t,I
0.5
0.4
0.3
-r 0.2
i,
% ALAs
45 40 35 30 25 20 15 10
- / I I I I 1 I /-
0.9_Lm LOC
--- 0.4/Lm LOC
m _
,, _, I _,, I [1,1 t I [ I,,, I, i J, I , , t 1
3.30 3.35 3.40 3.45 3.50 3.55 3.60
INDEX OF LOC LAYER (CSP-LOC)
>" 6Ol--
COZUJF--z 50H
r_COJWbJbju. rr 40
C.9rrW
bJ:z: 30I--
IJ.0
O.-1- 20
IJ.
% AL As
45 40 35 30 25 20 15 10
I 1 I I I I I im
m
m
m
!
m
m
m
m
n
m
//
/
//
//
//
/u
m
w
i
m
I
n
-,, ,,1 l,, ,I,, ,, I ,,,, I,, ,,1, ,, ,-
3.30 3.35 3.40 3.45 3.50 3.55 3.60
INDEX OF LOC LAYER (CSP-LOC) 6541
104
pAGE tS
OP,IGIHIkL QU,_L|'TYoi: poot_
#O
::LI0
I
105
o o
° o
cd
• .LISN3.LNI U7313 _V3N
0
IZ3
o
E::L
U3 X
d
00
6"3
d
0
I0
106
m
-o¢,.)
-- I !
- la.....I l_-_ 0'_ 0,') ¢.0_ _ i.t.i ¢.)
. I
7-111 I !!! I
o 80
d
($33_93C1)S3$VHd 0931-1 _lV3N
d
- 0O'7
m
m
i
m
u
m
m
0d
107
' I ' I ' I ' I00
0
0
Q._<_
I,iWn-(__ILl
I0
I I i I , I L I
d d d d
a o
W
I
I0d
111SN3±NI Q7313 _V3 03ZI7VB_ON
108
C)
t'q
O1¢}
i|ll j j j i i jll_Jj |Ill
\_ ,,,
- \\\ i
oo-!i iE E_-- :t. ::k '
-IIItwl I,,,,11"_,_
0 0 0
ii-i i:i:il0
yY
_.! --1
e-
O
(S33_30 )
AJ.ISN3.LNI 07313 _IV-I3HJ. 30 dHM-i
109
Z_ tO tO tO tO0
_9 o _ c_ 0a
_ o _o___
o___ ___
,,o
° N
C
w
m
I.-¢.D
110
111
-50 -40 -30
SIIBSTRATE
__L____J__
-20 -10
374-33
_L_L
0 10
8 (Degrees)
20 30 40 50
CAP
112
m
Z
I.--Zm
L_m
L.--
Lt--
m
._J
i
Z
0
m
p
m
I I I
-50SOBSTRATE
I
I
I I
0
8(Degrees)
374-159
I I
113
l==-i
ZLsJi===Zm
_.._sLs-Sm
Ls-.
i,t=lNm
......s
S
Z
0
I __
-50 0
374-28
50
-"--- SIIBSTRATE 8 (Degrees) CAP
114
m
ZL_
ZI
..I
m
WdN
m
E
0
!
0-50
--_- SUBSTRATE
0
8 (Degrees)
374-9
50
CAP
115
0cO
O m
O_
<_
O_c_J
u__J
O_ro
0O_
I0 0
(i. w0) S3SS07
0
116
0tOtO
"-... I :/ .. i -_/ ".. I -_./ "-. I - I_/ .. i -/ Z i -o° _/ . i - o &
OL .. , -- i_
/ - ' - _ "__>L ". ', _ <_
_F "- : _o_, dlore " , -- _ _:/-o " '. -_
/o_ % I.---, _ "'
_t- i ' _ - ,_ -__to ,,, ---: t_ _./. I o -, =4_l: I --= -i®_-IJ: I <" -114
I,, , ,, , ,:,'_,i-l°
t_ _ _. _. _. _.""
X3CINI3A1103 -I.-13
Appendix C
119
Effects of Ion Sputtering on the Surface Composition of GaAs Laser Diode Facets
S. E. Slavin, A. R. Triano, and L. A. DiMarco
RCA Laboratories, Princeton, NJ 08540
INTRODUCTION
In recent years ion sputtering has been investigated for application in various
phases of opto-electronic device fabrication processes 1,2. One principle interest
involves the creation of a clean and stoichiometric GaAs surface for subsequent
processing. Much of the previous work cited in the literature involves the use of some
combination of ion sputtering and thermal annealing to produce high quality surfaces 3-8.
Temperatures on the order of several hundred degrees celsius are reportedly necessary
to remove ion damage and recover surface stoichiometry 3"5,8. Ion sputtering without
any thermal treatment has generally resulted in deviations from stoichiometry 4-6,8-10
We have examined argon ion sputtering for the purpose of removing the native
oxide and any adsorbed impurities from the cleaved mirror facets of GaAs laser diodes
prior to the deposition of passivation coatings 11 Pre-deposition surface cleaning
should result in increased adhesion of the facet coatings and improved device reliability.
Unfortunately, the thermal sensitivity of the device structure at this advanced stage in the
fabrication process precludes the possibility of high temperature annealing to restructure
the surface after ion sputtering.
In the present study, we employ Auger spectroscopy to quantify changes in the
surface composition of (011) GaAs generated by argon ion bombardment in the absence
of any thermal annealing. Deviations from stoichiometry and surface oxide removal
rates are examined for argon ion current densities between 0.02 and 0.04 mA/sq.cm.
and ion energies in the range of 200 to 1000 eV.
_"_ //__ l_tP_r_,rt_,ir_ _ PRE(ZEI)[NG PAGE BLANK NOT FILMED
120
EXPERIMENTAL
The ion sputtering experiments were performed in a conventional Airco Temescal
cryopumped electron beam deposition system with a base pressure of 10-7 torr. The ion
source used was a broad beam type of the Kaufman design manufactured by the
Commonwealth Scientific Corporation. Physical constraints within the deposition
chamber resulted in placement of the ion gun at an angle of five degrees off normal to
the sample surface. The argon flow rate to the ion source was adjusted via a mass flow
controller to give an operating pressure of 10-4 torr during the ion sputtering process,
without cold head valve throttling. A biased plasma probe with a grounded shield could
be interchanged with the sample thus enabling in situ measurements of the ion current
density prior to actual sample exposure to the beam.
A gate valve isolates the deposition system from an attached analytical chamber
with a base pressure of 10"10 torr, containing a scanning Auger spectrometer (PHI
model 15-110B) and a secondary electron detector (PHI model 04-202). A magnetically
coupled linear-rotary feedthrough is used to transfer samples between the two systems,
enabling Auger analysis of the ion sputtered samples without atmospheric exposure. A
schematic diagram of the combined E-beam deposition and Auger analysis system is
shown in figure 1.
Samples used in this investigation were cleaved in atmosphere from AIGaAs
laser diode wafers grown by liquid phase epitaxy. Chemical cleaning was not employed
prior to loading samples into the deposition system for ion sputtering. Each sample was
examined in the analytical chamber via Auger spectroscopy prior to any ion sputtering.
Subsequently they were transferred back to the deposition system and ion sputtered
under a fixed set of conditions for a total time of 20 min. Ion sputtering was interrupted
121
ANALYSIS DEPOSITION
IonGun
CMA
IManipulator
loll
Pump
SED
NGateValve
IonGun
Crystal Opticalo
Monitor Monitor
Sample
E - Beam IHearth I
I Cryo-Pump
TransferArm
Figure 1 - Schematic diagram of the combined
E-beam deposition and Auger analysis system.
122
periodically to allow transfer of the sample back to the analytical chamber for
examination.
Auger spectra were excited with a 5 KeV electron beam at a beam current of 150
nA. Two regions of each sample were analyzed and the compositional information
averaged to obtain the gallium to arsenic ratios and oxide removal rates reported in this
manuscript. The largest observed variation in either the gallium or arsenic concentration
between two points on any single surface was one percent. Relative elemental
sensitivity factors for gallium and arsenic were determined empirically by comparing the
351 eV silver peak height from an ion sputtered silver sample to the appropriate LMM
gallium and arsenic peaks from in situ cleaved (011) GaAs samples run under identical
conditions. The gallium and arsenic concentrations varied from stoichiometry by a
maximum amount of 0.3 percent on these samples.
DISCUSSION
Figure 2(a) shows an Auger spectrum, taken prior to ion milling, from one of the
(011) GaAs facets that was cleaved in the atmosphere. Carbon and oxygen are the only
impurities observed to be present on the samples surface before ion sputtering.
Calculations indicate that the atomic ratio of gallium to arsenic on this surface is 1.03.
A spectrum from the same sample taken after 20 min of sputtering with 400 eV
argon ions at a current density of 0.03 mA/sq.cm, is shown in Figure 2(b). Examination of
this spectrum reveals that the carbon has been removed entirely, while trace amounts of
oxygen still remain on the sample's surface. The inability to completely remove oxygen
is likely due to adsorption of residual water in the deposition system prior to the transfer
operation. Trace amounts of argon are also observed to be present after sputtering. In
this particular case, calculations indicate that the argon ion bombardment has caused
123
dNdE
(a) !iii= ,= ,.
,' iiii,lJ!!!!! I
,,"iiiil!!!!!._
I I'l II I]11111
lillllI I
'_iii!I | ::: :i i i i i i, • , ,, |
iiiiii!!!!!!i i i i i i
: : , : : :
: : : ;: :
;;::::: : : : : :
!!!!!!: ' : : : I
iiiiiiiiiiii',I-,I,4,,-,I,,,,11,,,-I-
iiiiii.- : : : : :; : ; ; ; ;
i;ilii0 •lOO 400 604) _ 1000 1200 1400
iiII
ilI I
I1I !
I |
I1I I
I 1
I I
III!I I
i |I !
11
d.__.NdE
Ill| I
(b) '""I|llJ
Lilt;/111_I IIIi I ] i
I II111
I III
IIIIIII
llllll
ill"
!!!!!!||| I I _
III I I I ....
IIIII I IIIIlull,.,, I.IIIIIIll! IIIIII
troll !!!!!!lall I I I l I t
talJl !!!!!!lift I I
=" ;iillUI | | i
Q .=OO 400 609 800 1000 1200
I I
!
Electron Energy (eV)
Figure 2 - (a) Auger spectrum from an (011) GaAs sample cleaved in atmosphere, taken
prior to Ion sputtering. (b) Auger spectrum from the same sample after 20 rain. of
sputtering with 400 eV argon lens at a current density of 0.03 mA/cm 2.
124
the gallium to arsenic atomic ratio to increase to 1.2.
Gallium to arsenic ratios after 20 minutes of ion sputtering at two different current
densities are plotted against the argon ion energy in Figure 3. Examination of this figure
reveals that the magnitude of the gallium surface enrichment increases with increasing
argon ion energy over the entire range of ion energies investigated. The steady state
surface compositions, which were obtained for total sputtering times greater than 10 min,
did not appear to vary systematically with the ion current density.
Preferential sputtering in GaAs has been examined by several researchers under
a variety of ion sputtering conditions 9,10. Ion-energy dependent gallium surface
enrichment on (100) oriented GaAs, for ion energies above 600 eV, has been observed
previously12. The fact that no deviations from stoichiometry were seen at lower ion
energies was attributed to insufficient penetration of the argon ions into the GaAs lattice.
Calculations using the LSS method 13 imply that at energies below 600 eV the argon ion
range in gallium arsenide is on the order of the LMM electron escape depths 14.
Structural and electronic damage may well extend to even greater depths 15. We
therefore expect to observe deviations from stoichiometry at ion energies below 600 eV.
Our data suggests that gallium surface enrichment, whose magnitude is
proportional to the ion energy, is occuring at ion energies as low as 200 eV on (011 )
GaAs. Differences in the surface binding energy of gallium and arsenic, as shown by the
volitility of arsenic, should result in differences in the energy dependence of the partial
sputtering yields. It is therefore not surprising to observe ion-energy dependent surface
enrichment at energies as low as 200 eV, which is substantially above the sputtering
threshold (approximately 35 eV).
Figure 4 shows a plot of the normalized oxygen peak height vs. the ion sputtering
time for 1000 eV argon ions at three different ion current densities. The rate at which
125
._o
O_
1.50 -
1.40
1.30
1.20[] []
e,
[] 0.03 mA/sq, cm.
• 0.04 mA/sq, cm.
[]
1.10 . , , . , , ' ,
0.20 0.40 0.60 0.80 1.00 1.20
Ion Energy (KeV)
Figure 3 - Ga / As Ratio After 20 Min. of Ion Sputtering
126
"" 50.0
"1-
" 40.0
13.
'- 30.0_D
x0 20.0
•- lO.OEL_
0z 0.0
0.04 mA/sq, cm.
.-m 0.03 mA/sq, cm.
--_ 0.02 mA/sq, cm.
' i ' I ' I
0.0 10.0 20.0 30.0
Ion Sputtering Time (min.)
Figure 4 - Oxygen Removal Rates for 1000 eV Argon Ions.
127
oxygen is removed from the surface is clearly proportional to the ion current density as
expected. Similar results are observed at ion energies between 200 and 1000 eV.
An estimate of 10/_, for the native oxide thickness was determined based on the
sputter rates of AI20 3 and Ta20 5 standards. Using this thickness, oxide removal rates
of approximately 1,2,and 5 A/rain. were obtained for ion current densities of 0.02, 0.03,
and 0.04 mA/sq, cm. respectively. The tracking of the gallium and oxygen concentrations
during sputtering was suggestive of a gallium-rich mixed oxide. The oxide composition
and thickness estimates compare favorably with those observed previously 16
SUMMARY
Argon ion sputtering has been examined as a method of cleaning the cleaved
mirror facets of AIGaAs diode lasers prior to the application of passivation coatings.
Auger electron spectroscopy was employed to monitor the surface composition at
various intervals during the sputtering process. Prior to sputtering, the cleaved (011)
surfaces were found to contain carbon and oxygen. After sufficient ion sputtering to
obtain steady state surface compositions, trace amounts of argon and oxygen remained
on the surfaces. Quantitative analysis involving the magnitudes of the gallium and
arsenic LMM electron transitions revealed that the ion sputtered surfaces show evidence
of gallium enrichment whose magnitude appears to be proportional to the ion energy.
The oxygen removal rates, on the other hand, show a first order dependence on the
argon ion current density.
ACKNOWLEDGEMENT
The authors wish to express their gratitude to Drs. M. Ettenberg, J. H. Thomas III and C.
W. Magee for many useful discussions during the preparation of this manuscript.
128
REFERENCES
1W. Chen, L. M. Walpita, C. C. Sun, and W. S. C. Chang, J. Vac. Sci. Technol. B 4 (3),
701 (1986).
2G. A. Lincoln, M. W. Geis, S. Pang, and N. N. Efremow, J. Vac. Sci. Technol. B 1 (4),
1043 (1983)o
3j. Massies, P. Devoldere, and N. T. Linh, J. Vac. Sci. Technol. 15 (4), 1353 (1978).
4K. Jacobi and W. Ranke, J. Elect. Spect. and Rel. Phenom. 8,225 (1976).
5S. Sinharoy and R. A. Hoffman, IEEE Trans. Elect. Dev. 31 (8), 1090 (1984).
6p. Drathen, W. Ranke, and K. Jacobi, Surf. Sci. 77, L162 (1978).
7W. Ranke and K. Jacobi, Prog. Surf. Sci. 10, 1 (1981).
8p. Oelhafen, J. L. Freeouf, G. D. Pettit, and J. M. Woodall, J. Vac. Sci. Technol. B 1 (3),
787 (1983).
9Topics in Applied Physics, Vol. 52, Sputtering by Particle Bombardment II,
Edited by R. Behrisch, (Springer-Verlag, Berlin, 1983), pp. 73
101. L. Singer, J. S. Murday, and L. R. Cooper, J. Vac. Sci. Technol. 15 (2), 725 (1978).
11 I. Ladany, M. Ettenberg, H. F. Lockwood, and H. Kressel, Appl. Phy. Lett. 30,
87 (1977).
12A. van Oostrom, J. Vac. Sci. Technol. 13 (1), 224 (1976).
13j. Lindhard, M. Scharff, and H. E. Schiott, Kgl. Danske Videnskab. Selskab,
Mat-Fys. Medd. 33 (14), (1963).
14G. Ertl and J. Kuppers in: LOw Energy Electrons and Surface Chemistry,
(Chemie, Weinheim, 1974).
15y. Sekino, M. Owari, M. Kudo, and Y. Nehei, Jap. J. Appl. Phys. 25 (4), 538 (1986).
16S. p. Kowalczyk, J. R. Waldrop, and R. W. Grant, J. Vac. Sci. Technol. 19 (3),
611 (1981).
129
FIGURE CAPTIONS
Figure 1 - Schematic diagram of the combined E-beam deposition and Auger
analysis system.
Figure 2 - (a) Auger spectrum from an (011 ) GaAs sample cleaved in atmosphere, taken
prior to ion sputtering. (b) Auger spectrum from the same sample after 20 rain.
of sputtering with 400 eV argon ions at a current density of 0.03 mA/cm 2.
Figure 3 - GaJAs atomic ratio after 20 minutes of ion sputtering.
Figure 4 - Oxygen removal rates for 1000 eV argon ions.
Appendix D
9RECEDING PAGE BLANK NOT FILMED
133
Intrusions in the Active Layer of CSP Lasers
S. E. Slavin
F. Z. Hawrylo
J. J. Hughes
ABSTRACT
CSP laser diodes from various wafers were observed to develop a node or
dark spot in their near-fleld patterns after failure during lifetesting.
"Node" devices from each wafer were angle lapped and stained in an attempt
to uncover a physical mechanism for this common failure symptom. Optical
microscopic examination of the beveled cross sections revealed that an
intrusion, in the form of a spike at the tip of the zinc diffusion front,
had penetrated into the active layer of these diodes. The node observed in
the near-field pattern appears to be in a position corresponding to the
location of the spike in the active region.
In addition, cathodoluminescence measurements on other "node" diodes
revealed a dark line defect approximately 2 _m wide running parallel to and
in the middle of the 5 _m contact stripe. We believe that this dark line
defect is representative of a region of non-radiative recombination which
occurs in the portion of the active layer containing the intrusion.
Introduction
The diffusion of zinc into gallium arsenide and aluminum gallium
arsenide is an important part of the fabrication process of various
electronic and optoelectronic devices i-I0. In stripe geometry laser
diodes, the zinc diffused region provides additional current confinement and
improved ohmic contacting with the external metallization. Reliability
studies performed by various researchers II-14 have suggested that the
presence of zinc in some way contributes to the degradation of laser diodes.
Various mechanisms have been postulated in an attempt to explain this
phenomenon.
PRECFLD_O PAGE BLANK NOT FILMED
134
The present study was undertaken in response to the observation of a
common failure symptom which was seen during lifetesting of CSP laser diodes
from various wafers. These devices developed a node or dark spot in their
near-field patterns after lifetest failure. An example of a typical
pre-lifetest near-field pattern is shown twice on the left side of Figure
1(a). The right side of this figure shows the linear intensity scans of the
lateral and transverse mode patterns. The gaussian intensity profile is
characteristic of single spatial mode lasers. Figure l(b) shows the
post-lifetest near-field pattern and lateral mode pattern intensity scan
from the same laser diode. The reduced intensity at the center of the
pattern in Figure l(b) is the feature referred to as a node.
Several "node" devices from different _rafers were chosen for examina-
tion an attempt to isolate a single common physical feature which would
account for the observed node development.
Sample Preparation
A) Beveled Cross Sections
One degree angle lapping was the method chosen to prepare the samples
for examination via optical microscopy. This procedure produces a 57 X
magnification of the epilayer structure in one dimension. Figure 2 shows a
schematic representation (not to scale) of a laser diode and the direction
of the one degree bevel surface relative to the device structure.
Individual laser diodes are removed from their heat sinks, stripped of
their solder layers, and wax mounted on one edge of a steel angle block.
Gallium arsenide feedstock material is mounted adjacent to the device to
protect it during the polishing process. A piece of sapphire is affixed to
the opposite side of the block to provide the desired bevel angle. A
schematic representation of the angle block containing the mounted
components is seen in Figure 4. Angle lapping is done on a spiral-grooved
tin wheel charged with one micron diamond grit. The bevel angle is measured
with an autocollimator, which bounces a collimated beam of light off the
back surface of the angle block onto a scale calibrated in minutes of arc.
135ORIGINAL PAGE IS
OF POOR QUALITY
_um
Figure 1(a) - Pre-lifetest near-field patterns (left) and intensity scans I right)
5um
Figure I(b) - Post-lifetest near-field pattern (left) and intensitv scan iright)
136
Figure 2 - Schematic representation (not to scale) of a CSP laser diode
showing the one degree angle lap cross section plane.
137
ORIGINAL PAGE 1S
OF POOR QUALITY
Stripe
P-GaAs Cap Layer
P-AIGaAs Clad Layer
Active Layer
N-AIGaAs Clad Layer
Figure 3 - 500 X photomicrograph of a stained angle lapped
cross section of a good CSP laser diode
138
Sapphire I
I
iSteel Block
GaAs
F---2.._...J_Fee d s tock
_ Diode
Figure 4 - Schematic representation of the angle block structure
139
A two component "AB" etchant 15 is used to delineate the epilayer structure
on the beveled surface of the diode. The optical micrograph in Figure 3
shows the device structure revealed on the polished surface of a working
laser diode after staining.
B) Ca thodol uminescence Samples
Samples for cathodoluminescence analysis are prepared by removing the
device from the heat sink and stripping the contact metals from the
p-surface to expose the p-GaAs cap layer. Etching of the cap layer in the
zinc diffused stripe region is achieved using a K3FeCN6 -KOH etchant. Five
to ten seconds of exposure to the etchant is generally sufficient to
generate pits in the stripe which extend through the p-GaAs cap layer into
the p-AIGaAs clad layer. The differing aluminum contents of the various
epilayers enables the use of energy dispersive x-ray analysis to estimate
the location of the base of an etch pit in the laser diode structure.
Discussion
Shown in Figure 5 is a stained angle lapped cross section of a laser
diode which had developed a node in its near-field pattern after failure
during lifetesting. The two problematic features which can be observed in
this micrograph are the proximity of the zinc diffusion front to the active
layer and the spike which protrudes from the tip of the diffusion front into
the cavity. As seen in Figure 3 (good device), a separation distance
greater than 0.5 m is desired between the active layer and the zinc
diffusion front in the CSP laser structure. A beveled cross section of a
diode from another wafer which produced devices that failed via node
formation is shown in Figure 6. This particular device was never operated
prior to its being angle lapped and stained. The micrograph reveals that
prior to operation, this device possesses a zinc diffusion front that is
immediately adjacent to the active layer. The major structural difference
between the devices in Figures 5 and 6 is that the lifetested device
contains an intrusion in the active layer in the form of a spike at the tip
of the diffusion front. Examination of the post-life near-field patterns
140
ORIGINAL PAGE ISOF POOR QUALITY
Stripe
P-(;a.,\s Cap Laver
P AIGaAs Clad Laver
Channel
Actwe Laver
N AIGaAs Cla,:t Laver
Figure 5 • 500 X photomicrograph of a stained
angle lapped cross section of a failed CSP laser
diode containing a spike in the active laver
141
.0_ POOR (!t;AL/fy
....... Active Laver
N, A1GaA'- Clad Laver
Channel
Figure 6 - 500 X photomicrograph of an angle lapped
cross section of a CSP laser diode which was
not operated prior to beveling and staining
142
from several node devices reveals that the node which develops during
lifetesting appears to be in a position which corresponds to the location of
the spike in the active layer. This evidence suggests that the spike forms
during the operation of the laser diode and is involved in its degradation.
Supplementary information about this failure mode has been obtained via
cathodoluminescence analysis of diodes that developed a node after l ifetest
failure. The micrographs in Figure 7 show the corresponding secondary
electron image and the cathodoluminescence image of one such device. A dark
line defect is seen in the center of the stripe along its entire length in
the cathodoluminescence image. The dark line is most obvious in the stripe
etch pits and at the ends of the stripe. Energy dispersive x-ray analysis
revealed that aluminum is present at these locations, and we therefore
conclude that the etchant has opened up optical windows into the p-AiGaAs
clad layer in these areas. Little or no aluminum was detected outside of
the stripe region, indicating the continued presence of some portion of the
p-GaAs cap layer. A high-pass filter that transmits photons with
wavelengths greater than 780 nm was placed between the sample and the
detector to ensure that any observed radiation originated in the active
la yet.
A more highly magnified SEM micrograph of the largest etch pit towards
the center of the stripe is seen in Figure 8. A cathodoluminescence image
of this same etch pit and the adjacent one is shown in Figure 9. A dark
line defect which is in the center of the stripe and runs parallel to it is
observed in the pit. Absorption of photons originating in the active layer
by zinc in the stripe region above can be ruled out as the cause of the dark
line, since the dark line does not cover the full width of the zinc diffused
stripe. The position of the dark line in the stripe corresponds to the
location of the node in the near-field pattern.
E_K3E,0toc:
(.13
(-- I-.-,-
E:3
0"E}0c-
IX3
,0Cat_Ec:0L.
,.i,-)(,j,
i,.-.-.,
uJ
::,,.,.f,_
"0¢.-0¢j,
CO
143ORIGINAL PAGE IS
OF POOR QUALITY
_-+
L'
,_. C._
• - _2)
,.; L-3
k..J
_. ,>-t
:-- i---.-'
_J C_
_w.--
L,'? ,L"
'L_ _.
144
ORIGINAL PAGE IS
OF POOR QUALITY
Figure 8 - SEM micrograph of the largest etch pit in the
center of the stripe of the diode shown in figure 7
145
ORIGINAL PAGE IS
OF POOR QUALITY
Figure 9 - 3600 X cathodoluminescence image <)f the largest etch pit and
the adjacent etch pit in center of the stripe of tlle diode shown in figure 7
146
Summary
The fact that the position of a node in the near-field pattern
corresponds to the location of either a spike in the active layer or a dark
line defect in the stripe, strongly suggests that the three features are
related and involved in the degradation of these devices. Observation of
the node only after device failure in lifetesting implies that the spike and
the corresponding dark line defect are forming during the operation of the
diode. The exact composition and structure of the spike are not yet known
and are being investigated. The available evidence tends to indicate that
the spike contains zinc from the diffusion front and that it provides
non-radiative recombination sites which result in the observed dark line
defect. The heat generated during device operation may cause zinc migration
into the cavity when the diffusion front is sufficiently close to the active
Ia yer.
Ladany and Kressel II'16'17 suggest that a highly doped zinc diffused
region is a potential source of the defects that are involved in laser diode
degradation. They observed improved reliability in devices in which zinc
was eliminated from the active region. Our own preliminary lifetest data on
deep zinc diffused laser diodes implies a similar conclusion. Shallower
diffusions have been implemented since the discovery of the spikes in the
active layer and no nodal failures have been observed during lifetesting of
devices from wafers with the reduced zinc diffusion front depth.
Acknowl ed$ emen t
The authors would like to express their appreciation to Edwin R. Levin
for his assistante with the cathodoluminescence analysis and to Vito Rossi
for his aid in developing the angle lapping procedure.
References
147
l)
2)
3)
4)
J. J. Hsleh, "Zn Diffused, Stripe Geometry Double-Heterostructure
GainAsP/InP Diode Lasers", IEEE J. Quantum Electronics, Vol. QE-_5, pp.
694, August 1979.
G. Vassilieff and B. Saint-Cricq, "Zn Incorporation in GaAIAs Grown by
Liquid Phase Epitaxy and its Electrical Properties", J. Appl. Phys.
54(8), pp. 4581, August 1983.
J. R. Shealy and S. R. Chinn, "Simultaneous Diffusion of Zinc and
Indium into GaAs", Appl. Phy. Lett. 47(4), pp. 410, 15 August 1985.
Y. Yuan, K. Eda, G. A. Vawter, and J. L. Merz, "Open Tube Diffusion of
Zn into AiGaAs and GaAs", J. Appl. Phys. 54(I0), pp. 6045, October
1983.
5)
6)
7)
8)
9)
A. H. van Ommen, "Examination of Models for Zn Diffusion in GaAs", J.
Appl. Phys. 54(9), pp. 5055, September 1983.
S. E. Blum, M. B. Small, and D. Gupta, "Shallow Zn Diffusion in Liquid
Phase Epitaxial GaAs and GaAIAs at 600°C ", Appl. Phys. Lett. 42(i), pp.
108, 1 January 1983.
U. Gosele and F. Morehead, "Diffusion of Zinc into Gallium Arsenide",
J. Appl. Phys. 52(7), pp. 4617, July 1981.
S. K. Ageno, R. J. Roedel, N. Mellen, and J. S. Escher, "Diffusion of
Zinc into GaAIAs", Appl. Phy. Lett. (II), pp. 1193, I December 1983.
H. K. Choi and S. Wang, "GaAs/GaAiAs Deep Zinc-Diffused Channel
Substrate Laser", J. Appl. Phys. 54(6), pp. 3600, June 1983.
I0) S. B. Phatak, "An Open Tube Method of Zinc Diffusion in III-V
Compounds", IEEE Elect. Dev. Lett. 3(5), pp. [32, 5 May 1982.
Appendix E
PRECEDING PAGE BLANK NOT FILMED
151
A HI_;It-POWER CHANNELED-SUBSTRATE-PI,ANAR AIGaAs LASER*
B. Goldstein, M. EttenbergN. A. Dinke] and J. K. Butler
RCA Laboratories
Princeton, NJ 08540 (USA)
Abstract
A channeled-substrate-planar AIGaAs
operated to 190 mW cw.
(power-current curves,
behavior, far-field
laser has been
Principal optoelectronic behavior
single spatial and spectral mode
characteristics, modulation and
astigmatism properties) will be discussed. Evidence for
"hole-burning" is seen. Life-test data at power levels up
to 100 mW is presented.
We report the principal results to date of an ongoing study of high-power
CSP laser structures which, in our best case, has produced laser operation up
to 190 mW cw (as given by the power-current curve), single fundamental spatial
and spectral mode operation up to about 90 mW cw, with single spatial lobe
operation continuing to 150 mW; beyond 70 mW there are increasing line broad-
ening effects in the parallel far-field patterns accompanied by the appearance
and growth (in a minor way) of spectral sidebands. The basic laser structure,
grown by standard multi-bin LPE methods, is shown in Fig. l(a) in schematic
form, together with the pertinent dimensional and compositional information.
*This work was supported in part by NASA Langley Research Center, Hampton,
Virginia, under Contract No. NASI-17441.
_Southern Methodist University
Dallas, Texas 75275
PRECEDING PAGE I_LANK NOT FILMED
152
Figure l(b) is a photograph of a cleaved facet stained to delineate the
contours of the channel, the cladding layers and the cap. Of critical
importance is the avoidance of perturbations in the planarity of the active
layer over the channel, and means of achieving this will be discussed.
We show in Fig. 2 the power-output/current-input (P-I) curves, the
spectral content of the output, and the parallel and perpendicular far-field
patterns at different output power levels. The laser facets for these measure-
ments were coated with an AI203/Si dielectric stack to produce 90% reflectance
on the back facet, and an approximately _/4 A1203 layer to produce an ap-
proximately I0% reflectance on the front or emitting facet. The room tempera-
ture (23°C) cw threshold current is 48 mA and the differential quantum ef-
ficiency, q, at the emitting facet is 41%. The wavelength shift is that
expected from the bandgap shift due to joule heating and a 25°C/watt mounted-
diode thermal resistance. The full-widths-at-half-maximum (FWHM) at 20 mW for
the parallel and perpendicular far-field patterns are, respectively, 6.5 ° and
27 ° . It is worth noting that after failure at 190 mW cw the laser facet
visually showed no damage and the laser continued to be operable up to _I00-
120 mW cw.
In Fig. 3 we have plotted both the FWHM for the parallel far-field
pattern and the spectral side-band power (relative to the main spectral line)
as a function of operating cw power; both increase with power in a similar
way, including the almost identical slope changes at about 70 mW. Note that
the appearance of the far-field pattern continues to be that of a fundamental
mode (see inserts); and even though the asymmetry ultimately develops into
well-defined structure at 170 mW, nowhere is there any discontinuity in FWHM
one should expect if there had been any jump in spatial mode. These data
indicate the presence of a strong passive waveguide in the lateral direction
below 70 mW. However, starting at 70 mW"hole-burning" begins to distort the
dielectric profile, and any asymmetries in the now active waveguide at high
drive levels will produce concomitant distortions in the far-field patterns.
Typical laser modulation behavior is indicated in Fig. 4 where we show
the laser response to square current pulsesat 14_ duty cycle. The fall and
rise times are _.5 nsec (the limit of pulse resolution); note the almost
complete absence of tailing in both the leading and trailing edges of the
output pulse, as well as a minimum of ringing oscillation. Modulation
properties were found not to change at power levels up to 80 mW, the limit of
the experiment.
Laser astigmatism was determined by measuring the position of minimum
beam waist of the focussed far-field radiation. Such a plot is shown in Fig.
5 for both the horizontal and vertical directions, and the fact that the beam-
waist minima occur at the same point in both directions indicates no astigmatic
change with power level up to 75 mW. The relation between these results and
the nature of the guiding in the laser will be discussed.
Life-test results will be discussed, including those from operation at
accelerated modes at elevated temperatures (up to 500C) and powers (up to I00
"mW cw). Stability at 20 mW and 30°C is excellent, while even at the
accelerated modes it is good. One unit operated at 100 mW for over 700 hrs.
It is worth noting that as a rule laser failure is not accompanied by any
clear visual signs of external mirror facet damage.
Continuing research will be discussed with emphasis on channel shape and
the role of spatial hole-burning on high-power laser operation.
154
200
t,_oo
_z
G¢
5O
ie) (b)
FIG. I
CURRENT LASERPULSE PULSE
2 nHcldiv 2 nsec/div
.5 nsec/d iv 2 nsec/div
FIG. 4
(e)
_ n !
--.i/.
(b) (c)
t///"J1 emo -le _ .N
FIG.2
60roW
80row
x-x DIRECTIONo-y DIRECTION
3 x
:_L I l I I I I I I
25 mW
2_'_ 6_ _ _ 5OmW
:_x xoSxo2 o_1_3_ t_ 75roW
JJ-i ! n i i i = n ! i
e-POSITION
FIG. 5
b.IOO
ORIGINAL PAGE IS
OE POOR QUALITY
Appendix F
An
157
Efficient AIGaAs Channeled-Substrate-Planar
Distributed Feedback Laser
B. Goldstein a),G. Evans, J. Connolly, N. Dinkel, and J. Kirk
David Sarnoff Research Center
CN 5300
Princeton, N.J. 08853-5300
ABSTRACT
A wavelength-locked, AIGaAs channeled-substrate-planar
distributed feedback laser has been made that operates to 40 mW
pulsed. The Bragg grating is situaied at the shoulders of the
substrate channel and is sandwiched between two protective
layers of AIGaAs and GaAs. Overall power efficiencies of 15% have
been measured at 40 mW of output power.
a) Present address: Solarex Corporation, 826 Newtown-Yardley
Road, Newtown, Pa., 18940
PRECEDING PAGE BLANK NOT FILMED
159
An Efficient AIGaAs Channeled-Substrate-Planar
Distributed Feedback Laser
B. Goldstein, G. Evans, J. Connolly, N. Dinkel, and J. Kirk
David Sarnoff Research Center
CN 5300
Princeton, N.J. 08854-5300
The desirability of high power index-guided diode lasers
operating in a wavelength-locked single longitudinal mode and with
a well-defined lateral spatial mode is self-evident. Applications
to space communications, optical recording and fiber-optics
abound. The channeled substrate planar (CSP) AIGaAs laser1,2, 3
has by now been well-established as being routinely capable of
supplying all the above mentioned properties exce0t spectral
stabilization. Recently, efficient gain-guided distributed feedback
(DFB) AIGaAs lasers4, ridge-guide DFB lasers 5 grown by two-step
molecular beam epitaxial (MBE) processes, and buried
heterostructure DFB lasers 6 grown by three-step liquid phase
epitaxial (LPE) processes have been reported that operate at 8700-
88ooA.
In this letter we report a CSP-DFB laser operating at =8300A
with good overall efficiency that is grown by a one-step LPE
process on a substrate containing a second order (A=2415A) DFB
{aRECDDtNG PAGE BLANK NOR F1LMLD
160
grating. The grating is placed between two protective layers
grown by metalorganic chemical vapor deposition (MOCVD) with an
index step at the interface and is situated on the shoulders of the
substrate channel.
The insertion of the AI.15Ga.85As layer (in place of GaAs in a
conventional CSP laser) slightly increases (=15.6%) the magnitude
of the complex effective index step which controls the lateral
optical confinement: the real part of the complex effective
increases from 4.119x10 -3 to 5.999x10 -3, while the imaginary
part decreases from 4.40x10 -3 to 3.54x10 -3. Our calculations
show that the percentage of the lateral near-field intensity which
is outside the channel and interacts with the gratings ranges from
16% to 1% as the channel width increases from 2 _.m to 8 I_m for
the CSP parameters corresponding to Fig. 1. A CSP laser having a
channel width of 4.2 _m would have appoximately 4% of the lateral.
field in the grating region of the structure. Evanescent field
interaction with a grating to provide DFB has been previously
reported for a buried heterostructure laser7.
The structure is shown schematically in Fig. l(a) and in
stained cross-sectional cleave in Fig. l(b). Details of the device
can be seen more clearly in Fig. 1(c) where we show a stained
cross-section lapped at a 1" angle in the vertical direction. The
substrate is prepared by chemically-etching a second-order grating
using a photoresist mask obtained by standard holographic
161
interferometry into a 0.8 jJ.m thick n-type AI.15Ga.ssAs layer. A
0.12 p.m thick GaAs layer is then grown over the grating. These
layers are grown using MOCVD growth for improved uniformity and
surface morphology. MOCVD growth also improves the nucleation
of the GaAs layer on the AI.15Ga.85As grating surface. The
AI.15Ga.85As layer prevents meltback of both the vee-channel and
the grating while the GaAs layer grown over the grating, provides a
nucleating surface during subsequent LPE growth. The index step
at the grating interface is approximately 0.1.
After the grating and its protective layers have been
fabricated, a 4.2 _.m wide vee-channel is chemically etched into
the substrate and four layers are grown: 1) AI.33Ga.67As cladding
layer (0.3p.m, Nd=lxl018 cm3); 2)Alo6Ga.94As active layer
(0.07#m, Nd=lxl017 cm-3); 3) AI.33Ga.67As cladding layer (1.5p.m,
Na=Sxl017 cm-3); 4) GaAs capping layer (0.Tp.m, Nd=5xl017 cm-3).
The growth is performed at 800°C using a cooling rate of 1°C. This
procedure significantly reduces the complexity of fabricating DFB-
CSP structures and lends itself to a high-volume manufacturing
process. In addition, the current path in the device does not
include a regrowth interface which has been associated with
higher than normal series resistance for devices fabricated with
one or more growth interfaces in the current flow path 7. After LPE
growth, routine contacting technology completes the device which
includes Zn stripe diffusion for current-confinement (see Fig.
162
l(a)), Ti/Pt/Au for the p-contact and AuGe/Ni/Au for the n-
contact.
The CSP-DFB laser displayed DFB operation atpulsed (1% duty
cycle; 100 nsec) power levels up to 40 mW and at cw power levels
up to 10 mW. We show in Fig. 2, the change in the single
longitudinal mode spectrum for a DFB-CSP laser operating over the
temperature range of 24-32°C. The pulsed output power from the
device was maintained at 10 mW. The facets of the device were
coated (front-10%; rear-85%) and no attempt was made to suppress
the Fabry Perot modes. Sideband rejection ratios are in the range
of 18-23 db. The observed pulsed spectra were characteristic of
DFB lasers in that they remained single line for all temperatures 8.
A typical pulsed spectrum for this device operating at 28°C can be
seen in Fig. 2. The wavelength temperature dependence for both
pulsed and cw operation is about 0.7JL/°C. This behavior is that
expected for an AIGaAs laser of the given composition and layer
thickness operating at 8300 A9. The fact that the DFB behavior
described above occurs over a relatively small temperature range
and at powers only up to 40 mW pulsed can be associated with the
small grating depth (=100-200A) produced by chemical etching in
the dovetail direction. Improved performance is expected with
deeper (500 A to 1000 A) gratings.
Figure 3, (a) and (b) shows the P-I curve and far-field
patterns, respectively, demonstrating thresholds as low as 50 mA,
163
kink-free power curves and well-defined single spatial modes. The
overall efficiency of the laser at 40 mW (total input electrical
power divided by output optical power) is 15%. Thus,
incorporation of the grating and the accompanying extra layers (see
Fig. 1) do not interfere with the desirable spatial mode and high-
power properties of the basic CSP structure.
In conclusion, a channeled substrate planar laser with
distributed feedback has been fabricated using conventional
single-step liquid phase epitaxial growth. This DFB-CSP operated
in a stable single longitudinal mode up to 40 mW of output power
and displayed an overall power conversion efficiency of 15%.
The authors wish to express their appreciation to J. Berkshire,
D. Gilbert, M. Harvey, and D. Tarangioli for their technical
assistance and to M. Ettenberg for helpful discussions. This work
was supported, in part, by the National Aeronautics and Space
Administration, Langley, VA, under contract No. NAS1-17441.
164
Figure Captions
Figure 1. (a) Schematic diagram of CSP-DFB laser. (b) Stained
cross-sectional cleave of CSP-DFB structure. (c) Stained cross-
sectional Cleave lapped at a 1 ° angle in the vertical direction.
Note especially the beginning of meltback between the n-clad and
n-buffer layers.
Figure 2. Wavelength shift as a function of heat-sink
temperature for a CSP-DFB laser operating pulsed at an output
power of 10 roW.
Figure 3. (a) Power-current curves for a CSP-DFB laser. (b) Far-
field radiation patterns for a CSP-DFB laser.
166
References
1 T. Kuroda, M. Nakamura, K. Aiki, and J. Umeda, Applied Optics,
Vol 17, No. 20, 15 Oct. 1978, pp 3264-3267.
2 B. Goldstein, M. Ettenberg, N. Dinkel, and J. Butler, Appl. Phys.
LetL 47, 655 (1985).
3 T. Shibutani, M. Kume, K. Hamada, H. Shimizu, K. Itoh, G. Kano, and
I. Termoto, IEEE J. Quantum Electron. QE-23, 760 (1987).
4 K. Kojima, S. Noda, K. Mitsunaga, K. Kyuma, and T. Nakayama,
Technical Digest of 5th International Conference on Integrated
Optics and Optical Fibre Communication (International Institute
of Communications, Venice, Italy, 1985), p. 99.
5 S. Noda, K. Kojima, K. Mitsunaga, K. Kyuma, K. Hamanaka, and T.
Nakayama, IEEE J. Quantum Electron. QE-23,188 (1987).
6 y. Nakano and K. Tada, Appl. Phys. Lett. 49,1145 (1986).
7 W. Tsang, R. Logan, and L. Johnson, Appi. Phys. Lett. 34 (11), 752
(1979).
8 S. Noda, K. Kojima, K. Mitsunaga, K. Kyuma, and T. Nakayama,
Appl. Phys. Lett. 48 (1), 4 (1986).
166
9 K. Aiki, M. Nakamura, J. Umeda, A. Yariv, A. Katzir, and H. W. Hen,
Appl. Phys. Lett. 27, 145 (1975).
Q.030
_'_' •_e__ eL- ] _ _ _ _ ....
OF POOR QUALITY
167
L_
H-aI I \_I// I
0
I--
03Z
I
V
0_l.i.I
!
EL
168
0¢N
CO
0o
o._h.
¢o
oII
0
o0
¢N
0 _0
CO
A
0
IL_
!,_
¢)
E
V q_.6UOlOAeM0
169
&
0,W
4.h.
W>n"D0
>I ,.
IO.
0
Mw ' h_AAOd
0
N6SASea, ce _ral_l
Report Documentation Page
1. Reuort No.
NASA CR-4189
2. Government Accession No.
4. Title and Subtitle
High-Power AIGaAs Channeled Substrate Planar
Diode Lasers for S[_ceborne Conrmlnications
7. Author(sl
g. C. Connolly, B. Goldstein, G. N.
D. B. Carlin, and M. Ettenberg
Pultz, S. E. Slavin
9. Performing Organization Name and Address
David Sarnoff Research Center
Princeton, NJ 08543-5300
12. Sponsoring Agency Name and Address
National Aeronautics & Space Administration
Langley Research Center
Hampton, Virginia 23665-5225
3. Recipient's Catalog No.
5. Report Date
November 1988
6. Performing Organization Code
8. Performing Organization Re_ort No.
10. Work Unit No.
506-44-21-01
11. Contract or Grant No.
NASI-17441
13. Type of Report and Period Covered
Contractor Report8/4/86 through 7/15/87
14. Sponsoring Agency Code
15. Supplementary Notes
Langley Technical Monitor:
Final Report
Herbert D. Hendricks
16. Abst_ct
A high-power channeled substrate planar AIGaAs diode laser with an emission
wavelength of 8600-8800 _ has been developed. The opto-electronic behavior
(power-current, single spatial and spectral behavior, far-field characteristics,
modulation, and astigmatism properties) and results of con_utermodeling studies
on the performance of the laser will be discussed. Lifetest data on these devices
at high output power levels is also included. In addition, a new type of channeled
substrate planar laser utilizing a Bragg grating to stabilize the longitudinal
mode has been demonstrated. The fabrication procedures and optoelectronic
properties of this new diode laser are described.
17. Key Words (Suggested by Author(s))
semiconductor laser Bragg grating
high power distributed feedba
mode stabilized
!single longitudinal mode
!AIGaAs (Aluminum Gallium Arsenide)
19"SecuriwCIassif'(°fthisrep°n) 120" SecuriWClasmf"(°fthispage} 121"N°'°fpagesIUnc las s i f ied Unc ias s i f ied 168
NASA FORM 1_6 OCT
_r s_e by the N_ion_ Te_nic_ In_rmation Service, Sprin_eld, Virginia 22161-2171
18. Distribution Statement
Unclassified - Unlimited
ck Subject Cate_jory 36
22. Price
A08
NASA-Langley, 1988
1
i
I
i|i
N_ Postage end Fees Pa}dNational Aeronau'_tcs andSpace Administration
i Natlona! Aeronauticsand=_ Space Administration OmccatSugne.i
PextaJty for Pdvete UM $300Washington,D.C. SPECIALFOURTHCt.A_ MAIL2O546 BOOK